Find research datasets worth reusing
Search datasets from major research repositories and use ShareScore to quickly assess how well each record supports discovery, access, and reuse.
106
datasets available to search
ShareScore release 0.9.0
Dataset results
106 results for “Audio Dataset”
Dataset for the study Late development of audio-visual integration in the vertical plane
<p>It is not clear how multisensory skills develop and how visual experience impacts on multisensory spatial development. Conflicting results show that visual calibration precedes multisensory integration for the audio-visual spatial bisection task (Gori et al., 2012a, 2012b) while in other tasks such as spatial localization, visual calibration occurs after multisensory development (Rohlf et al., 2020). Results in blind individuals can say something about the role of vision on perceptual development. Scientific evidences show that blind individuals have impairments in bisecting the auditory space (Gori et al., 2014) but not in localizing auditory sources (Lessard et al., 1998). Such results suggest that sensory calibration and impairment are linked. We studied the development of audio-visual multisensory localization in the vertical plane in sighted individuals from 5 years to adulthood to address this hypothesis. We hypothesize that typical children would show late audio-visual integration for the vertical plane, preceded by visual dominance. Unimodal and bimodal audio-visual thresholds and PSEs were measured and compared with the Bayesian optimal-integration model (maximum likelihood estimation). Results show that the development of multisensory integration in the vertical plane is not evident at 5 years, suggesting visual dominance for vertical audio-visual localization. These results support the idea that multisensory perception in the vertical domain depends on sensory calibration. We discuss these scientific results proposing that the process of cross-sensory calibration is task-specific and highlighting the importance of linking the impairment and development to better determine how our brain works.</p> <p>Data are in textual tab delimited format. Columns report for each subject: age, age_bin, condition, jnd.</p> <p> </p>
PartialSpoof Database - Partially Spoofed Audio Dataset for Anti-spoofing
<p>All existing databases of spoofed speech contain attack data that is spoofed in its entirety. In practice, it is entirely plausible that successful attacks can be mounted with utterances that are only partially spoofed. By definition, partially-spoofed utterances contain a mix of both spoofed and bona fide segments, which will likely degrade the performance of countermeasures trained with entirely spoofed utterances. This hypothesis raises the obvious question: ‘Can we detect partially spoofed audio?’ This paper introduces a new database of partially-spoofed data, named <strong>PartialSpoof</strong>, to help address this question. This new database enables us to investigate and compare the performance of countermeasures on both utterance- and segmental- level labels. Experimental results using the utterance-level labels reveal that the reliability of countermeasures trained to detect fully-spoofed data is found to degrade substantially when tested with partially-spoofed data, whereas training on partially-spoofed data performs reliably in the case of both fully- and partially- spoofed utterances. Additional experiments using segmental-level labels show that spotting injected spoofed segments included in an utterance is a much more challenging task even if the latest countermeasure models are used.</p> <p> </p> <ul> <li><strong>!!!NEW!!! For detailed (bonafide/spoofing methods/nonspeech/concatenated parts) timestamps of PartialSpoof v1.3</strong> <ul> <li><a href="https://drive.google.com/drive/folders/1kKW3GBuooPkAl64Zyv6WPgICH5LZtnQR">Google Drive</a> </li> <li>The official version is under preparation. Please download this one if you urgently need it.</li> </ul> </li> <li>For fine-grained labels of PartialSpoof v1.2 <ul> <li>Arxiv: http://arxiv.org/abs/2204.05177</li> <li>PartialSpoof Database v1.2<strong> </strong>(including segmental-level labels in different temporal resolutions and timestamp labels)<strong>: This one</strong></li> </ul> </li> <li>For the multi-task version of PartialSpoof <strong>v1.1</strong> <ul> <li>Arxiv: https://arxiv.org/abs/2107.14132</li> <li>PartialSpoof Database v1.1 (including 0.16s segmental level labels): https://zenodo.org/record/5112031</li> </ul> </li> <li>For the initial version of PartialSpoof <strong>v1.0</strong> <ul> <li>Arxiv: https://arxiv.org/abs/2104.02518</li> <li>Samples: https://nii-yamagishilab.github.io/zlin-demo/IS2021/index.html</li> <li>PartialSpoof Database v1.0: https://zenodo.org/record/4817532</li> </ul> </li> </ul> <p>P.S.</p> <p>1. Compared to the <a href="../record/4817532#.YLO07S2l1hE">PartialSpoof_v1.0</a> and <a href="../record/5112031">PartialSpoof_v1.1</a>, only <strong>database_segment_labels_v1.2.tar.gz, database_vad.tar.gz, </strong> and<strong> README_v1.2</strong> are updated for version 1.2, you don't need to download other files if you already downloaded version1.0 or 1.1.</p> <p>2. File database_eval.tar.gz is a little large, if you cannot download it smoothly, you can download the split database_eval.tar.gz from <a href="../record/4817532#.YLO07S2l1hE">PartialSpoof_v1.0</a> </p>
Guinea baboon vocalizations dataset automatically extracted with a deep neural network from natural audio recordings
<p><strong>Abstract</strong></p> <p>The data collection process consisted of continuously recording during one month a group of Guinea baboons living in semi-liberty at the CNRS primatology center in Rousset-sur-Arc (France). Two microphones we placed nearby their enclosure to continuously record the sounds produced by the group. A convolutional neural network (CNN) was used on these large and noisy audio recordings to automatically extract segments of sound containing a baboon vocal production by following the method of <a href="https://arxiv.org/abs/2302.07640">Bonafos et al. (2023)</a>. The resulting dataset consists of one-second to several-minute wav files of automatically detected vocalizations segments. The dataset thus provides a wide range of baboon vocalizations produced at all times of the day. It can be used to study vocal productions of non-human primates, their repertoire, their distribution over the day, their frequency, and their heterogeneity. In addition to the analysis of animal communication, the dataset can also be used as a learning base for sound classification models.</p> <p> </p> <p><strong>Data acquisition</strong></p> <p>The data are audio recordings of baboons. The recordings were made with a H6 Zoom recorder, using the included XYH-6 stereo microphone. The sample size is 44100 Hertz, 16 bits. The microphones were placed in the vicinity of the enclosure for one month and recorded continuously on a PC computer. A CNN passed over the data with a sliding window of 1 second and an overlap of 80% to detect the vocal productions of the baboons. The dataset consists of the segments predicted by the CNN to contain a baboon vocalization. Windows containing signal less than one second apart were merged into a single vocalization.</p> <p> </p> <p><strong>Data source location</strong></p> <ul> <li>Institution: CNRS, Primate Facility</li> <li> <p>City/Town/Region: Rousset-sur-Arc</p> </li> <li> <p>Country: France</p> </li> <li> <p>Latitude and longitude for collected samples/data: 43.47033535251509, 5.6514732876668905</p> </li> </ul> <p> </p> <p><strong>Value of the data</strong></p> <ul> <li> <p>This dataset is relatively unique in terms of the quantity of vocalizations available.</p> </li> <li> <p>This massive dataset can be very useful to two types of scientific communities: experts in primatology who study the vocal productions of non-human primates, and experts in data science and audio signal processing.</p> </li> <li> <p>The machine learning research community has at its disposal a database of several dozen hours of animal vocalizations, which will make it possible to build up a large learning base, very useful for Environemental Sound Recognition tasks, for example.</p> </li> </ul> <p> </p> <p><strong>Objective</strong></p> <p>This dataset is a follow-up of two studies on the vocal productions of Guinea baboons (Papio papio) in which we carried out analyses of their vocal productions on the basis of a relatively large vocalization sample containing around 1300 vocalizations (<a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0169321">Boë, Berthommier, Legou, Captier, Kemp, Sawallis, Becker, Rey, & Fagot, 2017</a>; <a href="https://hal.science/hal-01649539">Kemp, Rey, Legou, Boë, Berthommier, Becker, & Fagot, 2017</a>). The aim was to collect a larger database using the technique of deep convolutional neural networks in order to 1) automatically detect vocal productions in a large continuous audio recording and 2) perform a categorization of these vocalizations on a more massive sample. A description of the pipeline that enabled these automatic detections and categorizations is given in <a href="https://arxiv.org/abs/2302.07640">Bonafos, Pudlo, Freyermuth, Legou, Fagot, Tronçon, & Rey (2023)</a>.</p> <p> </p> <p><strong>Data description</strong></p> <p>The data is a set of audio files in wav format. They are at least one second long (the size of the window), up to several minutes, if several windows are consecutively predicted as containing signal. Moreover, we add the labeled data we used to train the CNN which did the prediction. We also provide two hours of the continuous recordings to have an idea of the continuous recordings and test the code of the paper provided on <a href="https://gitlab.com/papers4375727/detection-and-classification-of-vocal-productions">gitlab</a>.</p> <p>In addition, there is a database in csv format listing all the vocalizations, the day and time of their production, and the prediction probabilities of the model.</p> <p> </p> <p><strong>Experimental design, materials and methods</strong></p> <p>The original recordings represent one month of continuous audio recording. Seven hours of this month were manually labelled. They were segmented and labelled according to whether or not there was a monkey vocalization (i.e., noise or vocalization) and, if there was a vocalization, according to the type of vocalization (6 possible classes: bark, copulation grunt, grunt, scream, yak, wahoo). These manually labelled data were used as a training set for a CNN, which was automatically trained following the pipeline of Bonafos et al. (2023). This model was then used to automatically detect and classify vocalization during the whole month of audio recording. It processes the data in the same way when predicting new data as it does when training. It uses a sliding window of one second with an overlap of 80%. It does not take into account information from previous predictions, but calculates the probability of a vocalization in each one-second window independently. It then iterates through the month. For each window, the model predicts two outputs: the probability that there is a vocalization and the probability of each class of vocalization.</p> <p>For the purpose of generating the wav files, if a window has a probability of a vocalization greater than 0.5, it is considered to contain a vocalization. If it is the first one, a vocalization is started at that moment. If the time windows that follow a vocalization also contain a vocalization, then the signal they contain is added to the first segment for which a vocalization has been detected. As soon as a one-second segment no longer contains a signal corresponding to a vocalization, the wav file is closed. If windows are predicted to contain no vocalizations, but are between two windows that contain vocalizations within 1 second of each other, then all windows are merged.</p>
An Open-set Recognition and Few-Shot Learning Dataset for Audio Event Classification in Domestic Environments
<p>The problem of training a deep neural network with a small set of positive samples is known as few-shot learning (FSL). It is widely known that traditional deep learning (DL) algorithms usually show very good performance when trained with large datasets. However, in many applications, it is not possible to obtain such a high number of samples. In the image domain, typical FSL applications are those related to face recognition. In the audio domain, music fraud or speaker recognition can be clearly benefited from FSL methods. This paper deals with the application of FSL to the detection of specific and intentional acoustic events given by different types of sound alarms, such as door bells or fire alarms, using a limited number of samples. These sounds typically occur in domestic environments where many events corresponding to a wide variety of sound classes take place. Therefore, the detection of such alarms in a practical scenario can be considered an open-set recognition (OSR) problem. To address the lack of a dedicated public dataset for audio FSL, researchers usually make modifications on other available datasets. This paper is aimed at providing the audio recognition community with a carefully annotated dataset for FSL and OSR comprised of 1360 clips from 34 classes divided into pattern sounds and unwanted sounds. To facilitate and promote research in this area, results with two baseline systems (one trained from scratch and another based on transfer learning), are presented.</p> <p> </p>
OrchideaSOL: an audio dataset of isolated musical notes, including mutes and extended playing techniques
<p>OrchideaSOL<br> ==========<br> Version 2.0, April 2020.<br> </p> <p> </p> <p>Created By<br> --------------</p> <p>Carmine-Emanuele Cella (1), Daniele Ghisi (1), Vincent Lostanlen (2), Fabien Lévy (3), Joshua Fineberg (4), Yan Maresz (5)<br> <br> (1): UC Berkeley<br> (2): New York University<br> (3): Columbia University<br> (4): Boston University<br> (5): Conservatoire de Paris</p> <p> </p> <p>Description<br> ---------------</p> <p><br> OrchideaSOL is a dataset of 13265 samples, each containing a single musical note from one of 14 different instruments:</p> <ol> <li>Bass Tuba</li> <li>French Horn</li> <li>Trombone</li> <li>Trumpet in C</li> <li>Accordion</li> <li>Contrabass</li> <li>Violin</li> <li>Viola</li> <li>Violoncello</li> <li>Bassoon</li> <li>Clarinet in B-flat</li> <li>Flute</li> <li>Oboe</li> <li>Alto Saxophone</li> </ol> <p> </p> <p>These sounds were originally recorded at Ircam in Paris (France) between 1996 and 1999, as part of a larger project named Studio On Line (SOL). One asset of OrchideaSOL is that it contains many combinations of mutes and extended playing techniques.<br> <br> The OrchideaSOL audio data can be used for creative purposes insofar at the use complies with the Ircam Forum License. Please visit: https://forum.ircam.fr/legal/contrat-de-licence-forum-ircam/</p> <p><br> The OrchideaSOL metadata can be used for creative purposes insofar at the use complies with the Creative Commons Attribution 4.0 International license (see below).<br> <br> OrchideaSOL can be used for education and research purposes. In particular, it can be employed as a dataset for training and/or evaluating music information retrieval (MIR) systems, for tasks such as instrument recognition, playing technique recognition, or fundamental frequency estimation. For this purpose, we provide an official 5-fold split of OrchideaSOL. This split has been carefully balanced in terms of instrumentation, pitch range, and dynamics. For the sake of research reproducibility, we encourage users of OrchideaSOL to adopt this split and report their results in terms of average performance across folds.</p> <p> </p> <p>Data Files<br> --------------</p> <p>OrchideaSOL contains 13265 audio clips as WAV files, sampled at 44.1 kHz, with a single channel (mono), at a bit depth of 16. This is equivalent to the audio quality of a compact disc. Audio clips vary in duration between two and ten seconds.</p> <p>Every audio file has a file path of the form:<br> <FAMILY>/<INSTRUMENT><+MUTE>/<TECHNIQUE>/<INSTR><+M>-<TECH>-<PITCH>-<DYN>-<INSTANCE>-<MISC>.wav</p> <p><br> where:</p> <ul> <li><FAMILY> corresponds to the instrument family: "Brass", "Keyboards" (includes accordion), "Strings", and "Winds" (i.e., woodwinds).</li> <li><INSTRUMENT> is the full name of the instrument.</li> <li><+MUTE> is the type of mute being used, such as "wah", "harmon", "piombo", or "sordina". If there is no mute, this field is absent.</li> <li><TECHNIQUE> is the type of playing technique.</li> <li><INSTR> is the abbreviation of the instrument.</li> <li><+M> is the abbreviation of the type of mute, if applicable.</li> <li><TECH> is the abbreviation of playing technique.</li> <li><PITCH> denotes the pitch of the musical note. This pitch is encoded in the American standard pitch notation: pitch class (C means "do") followed by pitch octave. According to this convention, A4 has a fundamental frequency of 440 Hz.</li> <li><DYN> denotes the intensity dynamics, ranked from pp (pianissimo) to ff (fortissimo).</li> <li><INSTANCE> contains additional information, when applicable. For example, for bowed string instruments, the same pitch may sometimes be achieved on different positions and different strings, resulting in small timbre differences. In this case the label "1c", "2c", "3c", or "4c" denotes the string which is being bowed. (The letter c originates from the word "corde", which means string in French.) By convention, the first string is the one with the highest pitch when played as an open string. Furthermore, on some wind instruments, the same note was played multiple times, e.g. at multiple durations. In this case, we use the label "alt1", "alt2", etc. to denote alternative instances of the note. If none of these tags apply, the <INSTANCE> field becomes "N", which stands for "Not Applicable".</li> <li><MISC> contains additional information, if applicable. In OrchideaSOL, some pitches were never recorded, and thus missing from the chromatic scale. In this case, the <MISC> tag contains a letter "R", to denote the fact that the corresponding WAV file has been obtained by transforming a different audio clip via some digital frequency transposition (similar to Auto-Tune). The letter "R" stands for "resampled". Furthermore, some pitches were slightly out of tune in comparison with the A440 tuning standard. Again, we applied some digital frequency transposition to correct them and put them exactly in tune. The amount of frequency transposition is measured in "cents" of an equal-tempered semitone. The letter "T" stands for "tuned". Because we employed a high-fidelity algorithm for frequency transposition, and because the amount of digital frequency transposition is small, the timbre of pitch-corrected notes remains faithful to the instrument. If none of these tags apply, the <MISC> field becomes "N", which stands for "natural"; in this case, the note is distributed exactly as it was recorded in the studio.</li> </ul> <p>For example, "Strings/Violin+sordina/tremolo/Vn+S-trem-A4-mf-4c-T13d_R200d.wav" corresponds to:</p> <ul> <li>a violin sound;</li> <li>equipped with a sordina mute;</li> <li>played in the tremolo playing technique;</li> <li>at pitch A4 (440 Hz);</li> <li>with mezzoforte dynamics;</li> <li>on the fourth string (i.e. the lowest);</li> <li>resampled from a B4 by lowering pitch by a semitone, i.e. 100 cents (R100d)</li> <li>lowered by 13 cents (T22d) to match the A440 tuning standard.</li> </ul> <p> </p> <p>The audio data for OrchideaSOL is not directly downloadable on Zenodo. Rather, it can be downloaded for free after registering to the Ircam forum. Please visit: https://forum.ircam.fr/</p> <p> </p> <p>Metadata File<br> -------------------</p> <p>The OrchideaSOL_metadata.csv file contains 13265 rows, one for each audio clip. It can be opened by a text editor or by a spreadsheet software application. It contains 13 columns:</p> <ol> <li>Path to the WAV file, in UNIX filesystem format. For Windows compatibility, replace the slashes ("/") by backslashes ("\"). Ex: "Strings/Violin+sordina/tremolo/Vn+S-trem-A4-mf-4c-T13d_R200d.wav"</li> <li>Fold ID. Either equal to 0, 1, 2, 3, or 4.</li> <li>Family. Ex: "Brass"</li> <li>Instrument abbreviation. Ex: "BTb"</li> <li>Instrument name in full. Ex: "Bass Tuba"</li> <li>Technique abbreviation.</li> <li>Technique name in full.</li> <li>Pitch. Ex: "A#1"</li> <li>Pitch ID in MIDI format. Ex: 34. Integer in the range 0-127.</li> <li>Dynamics. Ex: "ff".</li> <li>Dynamics ID. Integer. pp maps to 0 and ff maps to 4. The higher, the louder.</li> <li>Instance ID. Integer in the range 0-4</li> <li>String ID. Equal to 1, 2, 3, 4, or empty if not applicable.</li> <li>"Needed digital retuning". TRUE if the file has been pitch-shifted with digital audio effects; FALSE otherwise.</li> </ol> <p> </p> <p>Conditions of Use<br> ------------------------</p> <p>OrchideaSOL was created in 2020 by Carmine-Emanuele Cella, Daniele Ghisi, Vincent Lostanlen, Fabien Lévy, Joshua Fineberg, and Yan Maresz.</p> <p>OrchideaSOL is a derivative of SOL. We wish to thank Hugues Vinet, Greg Beller, and all coordinators of the Ircam Forum for their authorization to upload the metadata of OrchideaSOL to Zenodo.</p> <p>The audio samples in OrchideaSOL are offered free of charge under the Ircam Forum License. Please visit: https://forum.ircam.fr/legal/contrat-de-licence-forum-ircam/</p> <p>The dataset and its contents are made available on an "as is" basis and without warranties of any kind, including without limitation satisfactory quality and conformity, merchantability, fitness for a particular purpose, accuracy or completeness, or absence of errors. Subject to any liability that may not be excluded or limited by law, the authors are not liable for, and expressly exclude all liability for, loss or damage however and whenever caused to anyone by any use of the OrchideaSOL dataset or any part of it.</p> <p> </p> <p>Versions<br> -----------<br> 1.0 was released on February 24th, 2020.<br> 2.0 was released on April 4th, 2020. It fixes a bug in the instance IDs of oboe sounds in the "blow without reed" technique.</p> <p> </p> <p>Feedback<br> -------------</p> <p>Please help us improve OrchideaSOL by sending your feedback to:<br> carmine.cella@berkeley.edu</p> <p>For issues regarding the metadata encoding, the five-fold split, or the OrchideaSOL module in mirdata, please write to:<br> vincent.lostanlen@nyu.edu</p> <p>In case of a problem, please include as many details as possible.</p>
Dataset supplementing the article Einhäuser, W., Methfessel, P., & Bendixen, A. (2017). Newly acquired audio-visual associations bias perception in binocular rivalry. Vision Research, 133, 121-129.
<p>This dataset supplements the publication<br> Einhäuser, W., Methfessel, P., & Bendixen, A. (2017). Newly acquired audio-visual associations bias perception in binocular rivalry. Vision Research, 133, 121-129. doi: 10.1016/j.visres.2017.02.001</p> <p>Use is free for scientific purposes, provided the aforementioned reference is appropriately cited.<br> Description of files<br> - conditionsByObserver.csv<br> contains for each of the 16 observers the color and grating direction that had been coupled to either the low-pitch or the high-pitch tone<br> column 1: observer number<br> column 2: color associated with low-pitch tone<br> column 3: color associated with high-pitch tone<br> column 4: drift direction associated with low-pitch tone<br> column 5: drift direction associated with high-pitch tone</p> <p>- conditionsByObserver.mat contains the same information as matlab variables (as four vectors/cell arrays with one entry per observer)</p> <p>- toneByBlockAndTrial.csv<br> contains the conditions for all 18 rivalry trials (6 rivalry blocks with 3 trials each) for each observer<br> column 1: observer number<br> column 2: block number<br> column 3: trial number<br> column 4: tone (low [pitch], high [pitch], none) played in this trial<br> Note that due to a technical error for observer #16, block 6 was presented first, followed by 1,2,3,4,5; for all other observers blocks were used in the order given (1,2,3,4,5,6).</p> <p>- toneByBlockAndTrial.mat contains the same information as a 16x6x3 matrix named toneByBlockAndTrial ; tones are coded numerically (1-low pitch,2-high pitch,3-none)</p> <p>- eyeTraces.mat contains three cell arrays of dimensions 16x6x3 (observer x rivalry block x rivalry trial) called xEye, oknGain, and timeSinceTrialStart;</p> <p>o each entry of xEye contains the horizontal eye position for<br> the respective trial in eye-tracker coordinates (which correspond to screen pixels, except that (1/1) is the upper right rather than the upper left and values increase from right to left due to the setup configuration)</p> <p>o oknGain contains the gain computed from these eye positions.</p> <p>o timeSinceTrialStart contains the time in seconds since onset of the trial</p> <p><br> For all variables, the sampling rate is 500 Hz, in eye-tracker coordinates the speed of the grating is 240 units/ms. Blinks were removed from both eye-data variables, fast-phases were removed from the gain data. Removed data were set to NaN in eye-data variables.</p> <p>- Matlab functions figure1d.m, figure 2.m, figure3.m and figure4.m compute raw versions of the aforementioned paper's figures from the datafiles to exemplify their usage.</p> <p>[Note: In the originally published version of the article, the first two means and their standard errors of section 3.3 were stated incorrectly. All figures and statistical analyses are based on the correct data].</p>
WikiMuTe: A web-sourced dataset of semantic descriptions for music audio
<p>This upload contains the supplementary material for our <a href="https://arxiv.org/abs/2312.09207" target="_blank" rel="noopener">paper</a> presented at the <a href="https://mmm2024.org/" target="_blank" rel="noopener">MMM2024 conference</a>.</p> <h2>Dataset</h2> <p>The dataset contains rich text descriptions for music audio files collected from Wikipedia articles.</p> <p>The audio files are freely accessible and available for download through the URLs provided in the dataset.</p> <h3>Example</h3> <p>A few hand-picked, simplified examples of the dataset. </p> <table> <tbody> <tr> <td> <p><strong>file</strong></p> </td> <td> <p><strong>aspects</strong></p> </td> <td> <p><strong>sentences</strong></p> </td> </tr> <tr> <td> <p><a href="https://upload.wikimedia.org/wikipedia/commons/7/7a/Bongo_sound.wav" target="_blank" rel="noopener"><strong>🔈 Bongo sound.wav</strong></a></p> </td> <td> <p>['bongoes', 'percussion instrument', 'cumbia', 'drums']</p> </td> <td> <p>['a loop of bongoes playing a cumbia beat at 99 bpm']</p> </td> </tr> <tr> <td> <p><a href="https://upload.wikimedia.org/wikipedia/commons/4/46/Example_of_double_tracking_in_a_pop-rock_song_%283_guitar_tracks%29.ogg" target="_blank" rel="noopener"><strong>🔈 Example of double tracking in a pop-rock song (3 guitar tracks).ogg</strong></a></p> </td> <td> <p>['bass', 'rock', 'guitar music', 'guitar', 'pop', 'drums']</p> </td> <td> <p>['a pop-rock song']</p> </td> </tr> <tr> <td> <p><a href="https://upload.wikimedia.org/wikipedia/commons/6/62/OriginalDixielandJassBand-JazzMeBlues.ogg" target="_blank" rel="noopener"><strong>🔈 OriginalDixielandJassBand-JazzMeBlues.ogg</strong></a></p> </td> <td> <p>['jazz standard', 'instrumental', 'jazz music', 'jazz']</p> </td> <td> <p>['Considered to be a jazz standard', 'is an jazz composition']</p> </td> </tr> <tr> <td> <p><a href="https://upload.wikimedia.org/wikipedia/commons/5/58/Colin_Ross_-_Etherea.ogg" target="_blank" rel="noopener"><strong>🔈 Colin Ross - Etherea.ogg</strong></a></p> </td> <td> <p>['chirping birds', 'ambient percussion', 'new-age', 'flute', 'recorder', 'single instrument', 'woodwind']</p> </td> <td> <p>['features a single instrument with delayed echo, as well as ambient percussion and chirping birds', 'a new-age composition for recorder']</p> </td> </tr> <tr> <td> <p><a href="https://upload.wikimedia.org/wikipedia/commons/8/8b/Belau_rekid_%28instrumental%29.oga" target="_blank" rel="noopener"><strong>🔈 Belau rekid (instrumental).oga</strong></a></p> </td> <td> <p>['instrumental', 'brass band']</p> </td> <td> <p>['an instrumental brass band performance']</p> </td> </tr> <tr> <td> <p><strong>...</strong></p> </td> <td> <p>...</p> </td> <td> <p>...</p> </td> </tr> </tbody> </table> <h3>Dataset structure</h3> <p>We provide three variants of the dataset in the <code>data</code> folder.</p> <p>All are described in the paper.</p> <ol> <li><code>all.csv</code> contains all the data we collected, without any filtering.</li> <li><code>filtered_sf.csv</code> contains the data obtained using the <em>self-filtering</em> method.</li> <li><code>filtered_mc.csv</code> contains the data obtained using the <em>MusicCaps</em> dataset method.</li> </ol> <h3>File structure</h3> <p>Each CSV file contains the following columns:</p> <ul> <li><code>file</code>: the name of the audio file</li> <li><code>pageid</code>: the ID of the Wikipedia article where the text was collected from</li> <li><code>aspects</code>: the short-form (tag) description texts collected from the Wikipedia articles</li> <li><code>sentences</code>: the long-form (caption) description texts collected from the Wikipedia articles</li> <li><code>audio_url</code>: the URL of the audio file</li> <li><code>url</code>: the URL of the Wikipedia article where the text was collected from</li> </ul> <h3>Citation</h3> <div> <p>If you use this dataset in your research, please cite the following paper:</p> <div> <pre><code>@inproceedings{wikimute,</code><br><code> title = {WikiMuTe: {A} Web-Sourced Dataset of Semantic Descriptions for Music Audio},</code><br><code> author = {Weck, Benno and Kirchhoff, Holger and Grosche, Peter and Serra, Xavier},</code><br><code> booktitle = "MultiMedia Modeling",</code><br><code> year = "2024",</code><br><code> publisher = "Springer Nature Switzerland",</code><br><code> address = "Cham",</code><br><code> pages = "42--56",</code><br><code> doi = {10.1007/978-3-031-56435-2_4},</code><br><code> url = {https://doi.org/10.1007/978-3-031-56435-2_4},</code><br><code>}</code></pre> </div> </div> <h3>License</h3> <p>The data is available under the <a href="https://creativecommons.org/licenses/by-sa/3.0/" target="_blank" rel="noopener">Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) license</a>.</p> <p>Each entry in the dataset contains a URL linking to the article, where the text data was collected from.</p>
The UK COVID-19 Vocal Audio Dataset
<p>The UK COVID-19 Vocal Audio Dataset is designed for the training and evaluation of machine learning models that classify SARS-CoV-2 infection status or associated respiratory symptoms using vocal audio. The UK Health Security Agency recruited voluntary participants through the national Test and Trace programme and the REACT-1 survey in England from March 2021 to March 2022, during dominant transmission of the Alpha and Delta SARS-CoV-2 variants and some Omicron variant sublineages. Audio recordings of volitional coughs, exhalations, and speech (speech not available in open access version) were collected in the 'Speak up to help beat coronavirus' digital survey alongside demographic, self-reported symptom and respiratory condition data, and linked to SARS-CoV-2 test results. The UK COVID-19 Vocal Audio Dataset represents the largest collection of SARS-CoV-2 PCR-referenced audio recordings to date. PCR results were linked to 70,794 of 72,999 participants and 24,155 of 25,776 positive cases. Respiratory symptoms were reported by 45.62% of participants. This dataset has additional potential uses for bioacoustics research, with 11.30% participants reporting asthma, and 27.20% with linked influenza PCR test results.</p> <h3>Contents</h3> <ul> <li><strong>participant_metadata.csv</strong> row-wise, participant identifier indexed information on participant demographics and health status. Please see <a href="https://arxiv.org/pdf/2212.07738.pdf">A large-scale and PCR-referenced vocal audio dataset for COVID-19</a> for a full description of the dataset.</li> <li><strong>audio_metadata.csv</strong> row-wise, participant identifier indexed information on three recorded audio modalities, including audio filepaths. Please see <a href="https://arxiv.org/pdf/2212.07738.pdf">A large-scale and PCR-referenced vocal audio dataset for COVID-19</a> for a full description of the dataset.</li> <li><strong>train_test_splits.csv</strong> row-wise, participant identifier indexed information on train test splits for the following sets: 'Randomised' train and test set, Standard' train and test set, Matched' train and test sets, 'Longitudinal' test set and 'Matched Longitudinal' test set. Please see <a href="https://arxiv.org/abs/2212.08570">Audio-based AI classifiers show no evidence of improved COVID-19 screening over simple symptoms checkers</a> for a full description of the train test splits.</li> <li><strong>audio/ </strong>directory containing all the recordings in .wav format <ul> <li>Due to the large size of the dataset, to assist with ease of download, the audio files have been zipped into <strong>covid_data.z{ip, 01-24}.</strong> This enables the dataset to be downloaded in short periods, reducing the chances of a dropped internet connection scuppering progress. To unzip, first, ensure that all zip files are in the same directory. Then run the command 'unzip covid_data.zip' or right-click on 'covid_data.zip' and use a programme such as 'The Unarchiver' to open the file.</li> <li>Once extracted, to check the validity of the download, please run the 'python Turing-RSS-Health-Data-Lab-Biomedical-Acoustic-Markers/data-paper/unit-tests.py. All tests should pass with no exceptions. Please clone the GitHub repo detailed below.</li> </ul> </li> <li><strong>README.md</strong> full dataset descriptor.</li> <li><strong>DataDictionary_UKCOVID19VocalAudioDataset_OpenAccess.xlsx </strong>descriptor of each dataset attribute with the percentage coverage.</li> </ul> <h3>Code Base</h3> <p>The accompanying code can be found here: https://github.com/alan-turing-institute/Turing-RSS-Health-Data-Lab-Biomedical-Acoustic-Markers</p> <h3>Citations:</h3> <p>Please cite.</p> <p>@article{coppock2024audio,</p> <p> author = {Coppock, Harry and Nicholson, George and Kiskin, Ivan and Koutra, Vasiliki and Baker, Kieran and Budd, Jobie and Payne, Richard and Karoune, Emma and Hurley, David and Titcomb, Alexander and Egglestone, Sabrina and Cañadas, Ana Tendero and Butler, Lorraine and Jersakova, Radka and Mellor, Jonathon and Patel, Selina and Thornley, Tracey and Diggle, Peter and Richardson, Sylvia and Packham, Josef and Schuller, Björn W. and Pigoli, Davide and Gilmour, Steven and Roberts, Stephen and Holmes, Chris},</p> <p> title = {Audio-based AI classifiers show no evidence of improved COVID-19 screening over simple symptoms checkers},</p> <p> journal = {Nature Machine Intelligence},</p> <p> year = {2024},</p> <p> doi = {https://doi.org/10.1038/s42256-023-00773-8}</p> <p>}</p> <p>@article{budd2024,</p> <p> author={Jobie Budd and Kieran Baker and Emma Karoune and Harry Coppock and Selina Patel and Ana Tendero Cañadas and Alexander Titcomb and Richard Payne and David Hurley and Sabrina Egglestone and Lorraine Butler and George Nicholson and Ivan Kiskin and Vasiliki Koutra and Radka Jersakova and Peter Diggle and Sylvia Richardson and Bjoern Schuller and Steven Gilmour and Davide Pigoli and Stephen Roberts and Josef Packham Tracey Thornley Chris Holmes},</p> <p> title={A large-scale and PCR-referenced vocal audio dataset for COVID-19},</p> <p> journal={Scientific Data},</p> <p> year={2024},</p> <p> doi = {https://doi.org/10.1038/s41597-024-03492-w}</p> <p>}</p> <p>@article{Pigoli2022,</p> <p> author={Davide Pigoli and Kieran Baker and Jobie Budd and Lorraine Butler and Harry Coppock and Sabrina Egglestone and Steven G.\ Gilmour and Chris Holmes and David Hurley and Radka Jersakova and Ivan Kiskin and Vasiliki Koutra and George Nicholson and Joe Packham and Selina Patel and Richard Payne and Stephen J.\ Roberts and Bj\"{o}rn W.\ Schuller and Ana Tendero-Ca$\tilde{n}$adas and Tracey Thornley and Alexander Titcomb},</p> <p>title={Statistical Design and Analysis for Robust Machine Learning: A Case Study from Covid-19},</p> <p> year={2022},</p> <p> journal={arXiv},</p> <p> doi = {10.48550/ARXIV.2212.08571}</p> <p>}</p> <p> </p> <h3>The Dublin Core™ Metadata Initiative</h3> <p> </p> <p>- Title: The UK COVID-19 Vocal Audio Dataset, Open Access Edition.</p> <p>- Creator: The UK Health Security Agency (UKHSA) in collaboration with The Turing-RSS Health Data Lab.</p> <p>- Subject: COVID-19, Respiratory symptom, Other audio, Cough, Asthma, Influenza.</p> <p>- Description: The UK COVID-19 Vocal Audio Dataset Open Access Edition is designed for the training and evaluation of machine learning models that classify SARS-CoV-2 infection status or associated respiratory symptoms using vocal audio. The UK Health Security Agency recruited voluntary participants through the national Test and Trace programme and the REACT-1 survey in England from March 2021 to March 2022, during dominant transmission of the Alpha and Delta SARS-CoV-2 variants and some Omicron variant sublineages. Audio recordings of volitional coughs and exhalations were collected in the 'Speak up to help beat coronavirus' digital survey alongside demographic, self-reported symptom and respiratory condition data, and linked to SARS-CoV-2 test results. The UK COVID-19 Vocal Audio Dataset Open Access Edition represents the largest collection of SARS-CoV-2 PCR-referenced audio recordings to date. PCR results were linked to 70,794 of 72,999 participants and 24,155 of 25,776 positive cases. Respiratory symptoms were reported by 45.62% of participants. This dataset has additional potential uses for bioacoustics research, with 11.30% participants reporting asthma, and 27.20% with linked influenza PCR test results.</p> <p>- Publisher: The UK Health Security Agency (UKHSA).</p> <p>- Contributor: The UK Health Security Agency (UKHSA) and The Alan Turing Institute.</p> <p>- Date: 2021-03/2022-03</p> <p>- Type: Dataset</p> <p>- Format: Waveform Audio File Format audio/wave, Comma-separated values text/csv</p> <p>- Identifier: <strong>10.5281/zenodo.10043977</strong></p> <p>- Source: The UK COVID-19 Vocal Audio Dataset Protected Edition, accessed via application to <a href="https://www.gov.uk/government/publications/accessing-ukhsa-protected-data/accessing-ukhsa-protected-data">Accessing UKHSA protected data</a>.</p> <p>- Language: eng</p> <p>- Relation: The UK COVID-19 Vocal Audio Dataset Protected Edition, accessed via application to <a href="https://www.gov.uk/government/publications/accessing-ukhsa-protected-data/accessing-ukhsa-protected-data">Accessing UKHSA protected data</a>.</p> <p>- Coverage: United Kingdom, 2021-03/2022-03.</p> <p>- Rights: Open Government Licence version 3 (OGL v.3), © Crown Copyright UKHSA 2023.</p> <p>- accessRights: When you use this information under the Open Government Licence, you should include the following attribution: The UK COVID-19 Vocal Audio Dataset Open Access Edition, UK Health Security Agency, 2023, licensed under the <a href="https://www.nationalarchives.gov.uk/doc/open-government-licence/">Open Government Licence v3.0</a> and cite the papers detailed above.</p> <p> </p>
EEG and audio dataset for auditory attention decoding
<p>This dataset contains EEG recordings from 18 subjects listening to one of two competing speech audio streams. Continuous speech in trials of ~50 sec. was presented to normal hearing listeners in simulated rooms with different degrees of reverberation. Subjects were asked to attend one of two spatially separated speakers (one male, one female) and ignore the other. Repeated trials with presentation of a single talker were also recorded. The data were recorded in a double-walled soundproof booth at the Technical University of Denmark (DTU) using a 64-channel Biosemi system and digitized at a sampling rate of 512 Hz. Full details can be found in:</p> <ul> <li><strong>Søren A. Fuglsang, Torsten Dau & Jens Hjortkjær (2017): Noise-robust cortical tracking of attended speech in real-life environments. <em>NeuroImage</em>, 156, 435-444</strong></li> </ul> <p>and</p> <ul> <li><strong>Daniel D.E. Wong, Søren A. Fuglsang, Jens Hjortkjær, Enea Ceolini, Malcolm Slaney & Alain de Cheveigné: A Comparison of Temporal Response Function Estimation Methods for Auditory Attention Decoding. Frontiers in Neuroscience, </strong><a href="https://doi.org/10.3389/fnins.2018.00531">https://doi.org/10.3389/fnins.2018.00531</a></li> </ul> <p>The data is organized in format of the publicly available <a href="https://zenodo.org/record/1198430">COCOHA Matlab Toolbox</a>. The preproc_script.m demonstrates how to import and align the EEG and audio data. The script also demonstrates some EEG preprocessing steps as used the Wong et al. paper above. The AUDIO.zip contains wav-files with the speech audio used in the experiment. The EEG.zip contains MAT-files with the EEG/EOG data for each subject. The EEG/EOG data are found in <strong>data.eeg</strong> with the following channels:</p> <ul> <li>channels 1-64: scalp EEG electrodes</li> <li>channel 65: right mastoid electrode</li> <li>channel 66: left mastoid electrode</li> <li>channel 67: vertical EOG below right eye</li> <li>channel 68: horizontal EOG right eye</li> <li>channel 69: vertical EOG above right eye</li> <li>channel 70: vertical EOG below left eye</li> <li>channel 71: horizontal EOG left eye</li> <li>channel 72: vertical EOG above left eye</li> </ul> <p>The <strong>expinfo</strong> table contains information about experimental conditions, including what what speaker the listener was attending to in different trials. The expinfo table contains the following information:</p> <ul> <li>attend_mf: attended speaker (1=male, 2=female)</li> <li>attend_lr: spatial position of the attended speaker (1=left, 2=right)</li> <li>acoustic_condition: type of acoustic room (1= anechoic, 2= mild reverberation, 3= high reverberation, see Fuglsang et al. for details)</li> <li>n_speakers: number of speakers presented (1 or 2)</li> <li>wavfile_male: name of presented audio wav-file for the male speaker</li> <li>wavfile_female: name of presented audio wav-file for the female speaker (if any)</li> <li>trigger: trigger event value for each trial also found in data.event.eeg.value</li> </ul> <p>DATA_preproc.zip contains the preprocessed EEG and audio data as output from preproc_script.m.</p> <p>The dataset was created within the <a href="https://cocoha.org/">COCOHA</a><a href="https://cocoha.org/"> Project</a>: Cognitive Control of a Hearing Aid</p>
CrowdSpeech and Vox DIY: Benchmark Dataset for Crowdsourced Audio Transcription
<p>We collect and release CrowdSpeech — the first publicly available large-scale dataset of crowdsourced audio transcriptions. e show its applicability on an under-resourced language by constructing VoxDIY — a counterpart of CrowdSpeech for the Russian language.</p>
ESSENTIA analysis of audio snippets from the Million Song Dataset Taste Profile subset
<p>This upload includes the ESSENTIA analysis output of (a subset of) song snippets from the Million Song Dataset, namely those included in the Taste Profile subset. The audio snippets were collected from 7digital.com and were subsequently analyzed with ESSENTIA 2.1-beta3. Pre-trained SVM models provided by the ESSENTIA authors on their website were applied.</p> <p>The file <strong>msd_song_jsons.rar </strong>contains the ESSENTIA analysis output after applying the SVM models for highlevel feature extraction. Please note that these are 204317 files.</p> <p>The file <strong>msd_played_songs_essentia.csv.gz </strong>contains all one-dimensional real-valued fields of the jsons merged into one csv file with 204317 rows.</p> <p>The full procedure and subsequent analysis is described in</p> <p>Fricke, K. R., Greenberg, D. M., Rentfrow, P. J., & Herzberg, P. Y. (2019). Measuring musical preferences from listening behavior: Data from one million people and 200,000 songs. <em>Psychology of Music</em>, 0305735619868280.</p>
Dataset Interactive Audio Augmented Reality in Participatory Performance _Please Confirm You Are Not A Robot_
<p>This dataset gathers the different data from the study "Interactive Audio Augmented Reality in Participatory Performance". </p>
Audio Un-mixing Dataset (addendum)
<p>Listening test code and data supplementing the archive generated as part of the PhD research project "Improving the perceptual quality of single-channel blind audio source separation" by Toby Stokes.</p> <p>Further project details can be found at http://iosr.uk/unmixing</p> <p><strong>References</strong></p> <p>T.Stokes, "Improving the perceptual quality of single-channel blind audio source separation", PhD thesis, University of Surrey, UK, 2015</p>
Two datasets with user generated audio recordings
<p>We provide two open access datasets of <strong>user generated audio recordings </strong>captured with mobile devices such as smartphones and portable cameras . The provided audio files originate from two different public events, a <strong>musical concert</strong> and a <strong>football match</strong>. Also, for each event, we provide two different types of collections; the original <strong>unorganized</strong> collection of uncompressed audio files and an additional <strong>organized</strong> collection, where the different recordings corresponding to similar parts of the event are grouped into specific folders and time-aligned so they can be played back in unison. The interested researcher is invited to read the accompanying paper "<em>Two open access datasets of user generated audio recordings</em>" for finding out more details about these datasets and the way that they can be useful in the context of research related to the organization and reproduction of user generated content.</p>
StemGMD: A Large-Scale Audio Dataset of Isolated Drum Stems for Deep Drums Demixing - part 1
<p>We introduce StemGMD, a new large-scale dataset of isolated drum stems that builds upon the extensive MIDI collection found in <a href="https://magenta.tensorflow.org/datasets/groove">Magenta's Groove MIDI Dataset (GMD)</a>.</p> <p>GMD is a 13.6-hour corpus of expressive drum performances executed by ten drummers on a Roland TD-11 electronic drum kit. It contains 1150 MIDI files along with the corresponding full-kit audio mixtures.</p> <p>As a first step in creating StemGMD, we mapped the 22 different MIDI pitches found in the original files onto nine canonical instruments through the reduction scheme proposed in J. Gillick, A. Roberts, J. Engel, D. Eck, and D. Bamman, "Learning to groove with inverse sequence transformations," in International Conference on Machine Learning (ICML), vol. 97, 2019, pp. 2269–2279.</p> <p>Each of the nine resulting MIDI channels was manually synthesized as a 16-bit/44.1 kHz stereo WAV file using ten realistic-sounding acoustic drum kits sourced from the <a href="https://support.apple.com/en-me/guide/logicpro/lgsi2fb2509e/mac">Logic Pro X sample libraries</a>, i.e., Bluebird, Brooklyn, Detroit Garage, East Bay, Heavy, Motown Revisited, Portland, Retro Rock, Roots, and SoCal.</p> <p>As a result, StemGMD contains 1224 hours of audio, which correspond to more than 136 hours of full-kit mixtures. Moreover, StemGMD also contains single hits for each of the drum pieces at ten different velocities ranging from 30 to 127.</p> <p>To the best of our knowledge, StemGMD is the largest publicly available dataset of drums to date. Moreover, it is the first collection of single-instrument clips from all nine pieces in a canonical drum kit, making it well-suited for training deep drums demixing models.</p> <p> </p> <p><strong>*** THIS IS PART 1 OF 2 ***</strong></p> <p><strong>Download part 2 here:</strong> <a href="../records/7882857">https://zenodo.org/records/7882857 </a> (now available!) </p> <p>After downloading both parts, run <strong>unzip_StemGMD.sh</strong> to build the dataset from the split archive files.<br>Once unzipped, StemGMD will take just over <strong>1.13 TB </strong>of memory.</p> <p> </p> <p>The dataset is made available under a <a href="https://creativecommons.org/licenses/by/4.0/">Creative Commons Attribution 4.0 International (CC BY 4.0) License</a>.</p> <p>____________________________</p> <p>We employed the dataset in our paper titled "Toward Deep Drum Source Separation," published in Pattern Recognition Letters. </p> <div> <div> <div> <div> <p>Please, cite this work as: A. I. Mezza, R. Giampiccolo, A. Bernardini, and A. Sarti, "Toward Deep Drum Source Separation," Pattern Recognition Letters, vol. 183, pp. 86-91, 2024, doi: 10.1016/j.patrec.2024.04.026.</p> <pre>@article{mezza2024, title = {Toward deep drum source separation}, author = {Alessandro Ilic Mezza and Riccardo Giampiccolo and Alberto Bernardini and Augusto Sarti}, journal = {Pattern Recognition Letters}, volume = {183}, pages = {86-91}, year = {2024}, issn = {0167-8655}, doi = {https://doi.org/10.1016/j.patrec.2024.04.026} }</pre> </div> </div> </div> </div>
StemGMD: A Large-Scale Audio Dataset of Isolated Drum Stems for Deep Drums Demixing - part 2
<p>We introduce StemGMD, a new large-scale dataset of isolated drum stems that builds upon the extensive MIDI collection found in <a href="https://magenta.tensorflow.org/datasets/groove">Magenta's Groove MIDI Dataset (GMD)</a>.</p> <p>GMD is a 13.6-hour corpus of expressive drum performances executed by ten drummers on a Roland TD-11 electronic drum kit. It contains 1150 MIDI files along with the corresponding full-kit audio mixtures.</p> <p>As a first step in creating StemGMD, we mapped the 22 different MIDI pitches found in the original files onto nine canonical instruments through the reduction scheme proposed in J. Gillick, A. Roberts, J. Engel, D. Eck, and D. Bamman, "Learning to groove with inverse sequence transformations," in International Conference on Machine Learning (ICML), vol. 97, 2019, pp. 2269–2279.</p> <p>Each of the nine resulting MIDI channels was manually synthesized as a 16-bit/44.1 kHz stereo WAV file using ten realistic-sounding acoustic drum kits sourced from the <a href="https://support.apple.com/en-me/guide/logicpro/lgsi2fb2509e/mac">Logic Pro X sample libraries</a>, i.e., Bluebird, Brooklyn, Detroit Garage, East Bay, Heavy, Motown Revisited, Portland, Retro Rock, Roots, and SoCal.</p> <p>As a result, StemGMD contains 1224 hours of audio, which correspond to more than 136 hours of full-kit mixtures. Moreover, StemGMD also contains single hits for each of the drum pieces at ten different velocities ranging from 30 to 127.</p> <p>To the best of our knowledge, StemGMD is the largest publicly available dataset of drums to date. Moreover, it is the first collection of single-instrument clips from all nine pieces in a canonical drum kit, making it well-suited for training deep drums demixing models.</p> <p> </p> <p><strong>*** THIS IS PART 2 OF 2 ***</strong></p> <p><strong>Download part 1 here:</strong> <a href="../records/7860223">https://zenodo.org/records/7860223 </a></p> <p>After downloading both parts, run <strong>unzip_StemGMD.sh</strong> to build the dataset from the split archive files.<br>Once unzipped, StemGMD will take just over <strong>1.13 TB </strong>of memory.</p> <p> </p> <p>The dataset is made available under a <a href="https://creativecommons.org/licenses/by/4.0/">Creative Commons Attribution 4.0 International (CC BY 4.0) License</a>.<br><br>_____________________________<br><br>We employed the dataset in our paper titled "Toward Deep Drum Source Separation," published in Pattern Recognition Letters. </p> <div> <div> <div> <div> <p>Please, cite this work as: A. I. Mezza, R. Giampiccolo, A. Bernardini, and A. Sarti, "Toward Deep Drum Source Separation," Pattern Recognition Letters, vol. 183, pp. 86-91, 2024, doi: 10.1016/j.patrec.2024.04.026.</p> <pre>@article{mezza2024, title = {Toward deep drum source separation}, author = {Alessandro Ilic Mezza and Riccardo Giampiccolo and Alberto Bernardini and Augusto Sarti}, journal = {Pattern Recognition Letters}, volume = {183}, pages = {86-91}, year = {2024}, issn = {0167-8655}, doi = {https://doi.org/10.1016/j.patrec.2024.04.026} }</pre> </div> </div> </div> </div>
Extracting Biomedical Entities from Noisy Audio Transcripts--Dataset
<p><strong>SUMMARY</strong>:</p> <p>This repo contains the CADEC and Synthetic BTACT datasets that were used for the paper titled "<em>Extracting Biomedical Entities from Noisy Audio Transcripts</em>."</p> <p>The dataset includes two sets: i) CADEC (Karimi et al., 2015) and ii) Synthetic BTACT. CADEC is a well-known NER dataset used to identify adverse drug reactions based on what patients have written about their experiences. Synthetic BTACT is the data that we have made up. It is created based on questions similar to those in the Brief Test of Adult Cognition by Telephone (BTACT)(Tun et al., 2006).</p> <p>CADEC includes two sets of audio files; one is read from the original CADEC, and the other one is with additional audio noise. It also includes the original CADEC scripts, annotations, and the transcripts of the noisy audio. The transcripts are generated using Whisper. The annotations encompass named entities, their types, and string indexes of their occurrence in the text. Annotations also include "AnnotatorNotes" which explains some of the annotations.</p> <p>The synthetic BTACT data include two types: i) animals and ii) fruits. Similar to CADEC, it includes two sets of audio files: one that is read from the original scripts and another one with additional audio noise. The text files include the original scripts, annotations, and the Whisper-transcribed of the noisy audio files. The annotations include indexes of named entities, their string indices and types.</p> <p><strong>REFERENCES</strong>:</p> <p>Karimi, S., Metke-Jimenez, A., Kemp, M., & Wang, C. (2015). Cadec: A corpus of adverse drug event annotations. Journal of biomedical informatics, 55, 73-81.</p> <p>Tun, P. A., & Lachman, M. E. (2006). Telephone assessment of cognitive function in adulthood: the Brief Test of Adult Cognition by Telephone. Age and Ageing, 35(6), 629-632.</p> <p><strong>DETAILS</strong>:</p> <p>Data_1: CADEC (1250 TextFiles, 1000 Audio, types=5):</p> <p> General Categories and Counts<br> ADR (Adverse Drug Reactions): 5316<br> DRUG: 1797<br> FINDING: 397<br> DISEASE: 280<br> SYMPTOM: 255<br> Specific Items (Drugs) and Counts<br> Arthrotec: 145<br> cambia: 4<br> cataflam: 10<br> diclofenac-potassium: 3<br> diclofenac-sodium: 7<br> flector: 1<br> Lipitor: 997<br> Pennsaid: 4<br> solarez: 3<br> voltaren: 46<br> voltaren-rx: 22<br> zipsor: 5</p> <p>Data_2: Synthetic BTACT (500 Fruits, 500 Animals, types=2)</p> <p>>> Audios can be matched with annotations, scripts and transcripts using their filenames. </p> <p>---<br>audio [original&noisy]:<br> 1. cadec<br> 1.1 cadec original<br> 1.2 cadec noisy<br> 2. synthetic btact<br> 2.1 btact original<br> 2.1.1 fruits<br> fruit-script[0:500].mp3<br> 2.1.2 animals<br> script-[0:500].mp3<br> 2.2 btact noisy<br> 2.2.1 fruits<br> fruit-script[0:500].mp3<br> 2.2.2 animals<br> script-[0:500].mp3<br>text[scripts, annotations, transcripts]:<br> 1. cadec<br> 1.1 scripts [1,250]<br> 1.2 annotations [1,250] (index/AnnotatorsNote, type, indices, named-entities)<br> 1.3 transcripts [1,000]<br> 2. synthetic btact<br> 2.1 animals<br> 2.1 scripts (original scripts)<br> script-[0:500].txt<br> 2.2 annotations<br> script-[0:500].ann (index, type, start/end indices, named entity)<br> 2.3 transcripts<br> script-[0:500].txt<br> 2.2. fruits<br> 2.1 scripts (original scripts)<br> script-[0:500].txt<br> 2.2 annotations<br> script-[0:500].ann (index, type, start/end indices, named entity)<br> 2.3 transcripts<br> fruit-script-[0:500].txt</p> <p> </p> <p><strong>CITATION</strong>:</p> <p>Ebadi, N., Morgan, K., Tan, A., Linares, B., Osborn, S., Majors, E., Davis, J., & Rios, A. (2024). Extracting biomedical entities from noisy audio transcripts. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024).</p>
AAM: Artificial Audio Multitracks Dataset
<p>This dataset contains 3,000 artificial music audio tracks with rich annotations. It is based on real instrument samples and generated by algorithmic composition with respect to music theory.</p> <p>It provides full mixes of the songs as well as single instrument tracks. The midis used for generation are also available. The annotation files include: Onsets, Pitches, Instruments, Keys, Tempos, Segments, Melody instrument, Beats, and Chords.</p> <p>A <strong>presentation</strong> <strong>paper</strong> was published open-access in <a href="https://doi.org/10.1186/s13636-023-00278-7">EURASIP Journal on Audio, Speech, and Music Processing</a>.</p> <p>Current development<strong> </strong>and source code of the <strong>generator tool</strong> can be found on <a href="https://github.com/fabianostermann/ArtificialSongGenerator">GitHub</a>.</p> <p>For a <strong>tiny version</strong> for demonstration and testing purposes see: <a href="https://doi.org/10.5281/zenodo.6771120">zenodo.6771120</a></p>
A dataset of environmental audio recordings containing chainsaw events
<p>This audio dataset contains audio recordings (.wav format, 8kHz sampling rate) of various durations acquired using eight Cornel University SWIFT Autonomous Recording Units (ARUs) and corresponding metadata (.textgrid files). The recordings were collected within the Rodopi Mountain-Range National Park, Greece, at different seasons between 2018 and 2019. They are part of longer audio recordings. This audio content was used for the evaluation of a chainsaw sound detection system that was developed for the needs of a research project funded via a Single RTDI State Aid Action ”Research – Create – Innovate” grant (T1EDK-04488), which is co-financed by Greece and the European Union (European Regional Development Fund) as part of the Operational Program ”Competitiveness, Entrepreneurship and Innovation” of the National Strategic Reference Framework (NSRF) 2014-2020.</p> <p>Portions of each uploaded recording contain chainsaw events. The temporal location of these events, originally located manually by human listeners, are marked within the corresponding .textgrid files. Each audio recording with the corresponding textgrid file can be opened using the Praat program (<a href="https://www.fon.hum.uva.nl/praat/">https://www.fon.hum.uva.nl/praat/</a>).</p> <p>If you find this dataset useful please cite the following paper:</p> <p>N. Stefanakis, K. Psaroulakis, N. Simou and C. Astaras, "An open-access system for long-range chainsaw sound detection", in Proceedings of EUSIPCO (2022).</p> <p>The chainsaw sound detection algorithm that was developed is also open access and is available in the form of python code from <a href="https://github.com/spl-icsforth/An-open-access-system-for-long-range-chainsaw-sound-detection">https://github.com/spl-icsforth/An-open-access-system-for-long-range-chainsaw-sound-detection</a></p>
Audio problems dataset
<p>Audio problems dataset for academic purposes.</p>
ScienceDex guides
Understand access before you commit
These curated guides explain access requirements, typical timelines, costs, and reuse considerations for widely used research datasets.
Allen Brain Atlas
Allen Brain Atlas is an Allen Institute collection of brain map atlases, datasets, APIs, and analysis tools covering mouse, human, and non-human primate brain resources.
Annotated Behaviour and Observability Dataset (ABODe)
ABODe is a University of Edinburgh DataShare dataset for behavior classification in group-housed mice using home-cage video, identities, bounding boxes, ground-plate positions, and annotator labels.
DANDI Archive for NWB datasets
DANDI is a BRAIN Initiative archive for publishing and sharing neurophysiology data, including electrophysiology, optophysiology, and behavioral data packaged as NWB and related standards.
International Brain Laboratory public data
The International Brain Laboratory public data releases expose standardized mouse decision-making experiments, including Neuropixels recordings, widefield calcium imaging, behavior, and session metadata accessed through the ONE API.
OpenNeuro
OpenNeuro is a free, open platform for sharing neuroimaging datasets, with public search, dataset pages, and download paths for web, S3, DataLad, and the OpenNeuro CLI.