Find research datasets worth reusing

opencc-by-4.0Aug 2015View details →

Figure 3. Electrode placement diagram-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

Reference electrode placed AFz placed in between AF1 and AF2 electrode and ground electrode Oz is placed between O1 and O2 electrodes. The impedance of the electrode range is 5KΩ. The sampling rate was fixed range between 256 samples per second for all the channels. Figure 3 shown in below which is represent the electrode placement diagram. The recorded EEG signal is used to recognize the different level of emotions.

opencc-by-4.0Aug 2015View details →

Figure 2. Emotion recognition using EEG-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

This section describes that collection of EEG signals for different emotion recognition experiments. The electroencephalography signals of 32 participants were recorded during one minute videos. Based on that participants are rated in terms of valence and arousal, like/dislike, familiarities and dominance. Emotion related ratings are given based on the online self assessment which is 120 one minute extracted music videos, which are rated by 14-16 volunteers based on arousal and valence. The EEG signals were recorded using 64 electrodes, first 62 electrodes are active electrode, one for reference and remaining one is ground electrode. All the electrodes are placed on the scalp which is made up of the Ag/Ag-Cl. Figure 2 shown in below which is represent the basic EEG signal recording methods and emotion analysis process.

opencc-by-4.0Aug 2015View details →

Figure 1. Brain Structure-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

EEG data have collected from desirable subjects. Each and every EEG signal has different kind of bands like Alpha, Beta, Gamma, Theta, and Delta. Each band stores the particular information about the emotions. Alpha band (8-13 Hz) which located in Frontal Occipital, Beta band (13-30 Hz) which located in Frontal Central, Gamma band (30-100 Hz), Theta band (4- 7 Hz) which located in Midline Temp, Delta band (0-4Hz) which located in Frontal Lobe. Before processing the EEG signal and extracting these bands, preprocess the signal and reduce the noise. The basic brain figure is shown in below.

Figure 8. Sample Emotion Classification Result-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

For all the 32 participants the EEG signal has to be sampled and process their emotions. The emotions are depends on the music and video clips. In this paper the video clips are changed from one person to other person. Here hybrid feed forward neural networks with radial basis function; probabilistic neural network classifier is used to classify the emotions from EEG. PNN is very fast and insensitive neural network which provides the optimized classification result. Compare to the multi layer perception neural network it provide accurate result. It classifies the emotion into two different groups like arousal and valence. Figure 8 shows that the model implemented result of emotion classification and person identification.

Figure 7. Neural Network model-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

In Probabilistic Neural Network the operations are organized into a multilayer feed forward neural network with four layers like input layer, hidden layer, pattern layer and output layer. PNN use the Euclidean distance measure the difference between one neuron to other neurons. The actual target values are stored in the hidden neuron and the optimized weighted values are fed into the same category hidden neuron. Then finally the output layer compared the weighted votes of each target values and the target votes are used to predict the emotions.

Figure 10. Sensitivity and specificity of different bands-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

Figure10 has shown in sensitivity and specificity of different neural network which is used to explain how exactly the emotions are classified into groups and accuracy value shown in above table 3.

Figure 9. Mean Square Error for different bands-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

Figure 9 has shown in different epochs using neural networks with mean square error performance and the Table 3 shows that different bands mean square error values while training the neural networks with particle swarm optimization.

Figure 6. Mean square error of alpha band-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

Particle swarm optimization algorithm first optimizes the neural networks weight and bias and provides the minimum mean square error with nearer by zero. The following figure 6 has shown that minimum mean square error when training the particular band features.

Figure 5. Process flow of PSO-Classification of Human Emotion from Deap EEG Signal Using Hybrid Improved Neural Networks with Cuckoo Search

PSO is used to identify the best solution from collection of solution. It is a computational method that optimizes a problem by iteratively trying to improve a candidate solution with regard to a given measure of quality. PSO optimizes a problem by having a population of candidate solutions, here dubbed particles, and moving these particles around in the search-space according to simple mathematical formulae over the particle's position and velocity. Each particle's movement is influenced by its local best known position but, is also guided toward the best known positions in the search-space, which are updated as better positions are found by other particles. This is expected to move the swarm toward the best solutions. PSO is a metaheuristic as it makes few or no assumptions about the problem being optimized and can search very large spaces of candidate solutions. However, metaheuristic such as PSO do not guarantee an optimal solution is ever found. The following Figure 5 explains the basic flow of PSO process.

opencc-by-4.0Jan 2016View details →

Figure 1. CNN architecture (adopted from Krizhevsky et al. '12)-Measuring Customer Behavior with Deep Convolutional Neural Networks

The architecture of a CNN can be described as following. A small pixel region goes to input neurons and then connects to a first convolution hidden layer (Figure1). There we can see a set of learnable filters, which are activated during the presentation some particular type of feature in pixel region in the input. On this phase, CNN does shift invariance, which is carried by feature map. Subsampling layer goes next. There we have two processes: local averaging and sampling. As a result, we get declining resolution of feature map. To correspond this task CNN needs supervised learning. Before starting the experiment, we gave a set of labeled videos with different emotional experience. The system analyses images and finds similar features. Then the system creates a map, where it arranges videos in accordance with similar features. Thereby, images with similar emotions form certain class.  To test the system, we add other videos and correct the system when it refers them improperly. The proposed model consists of four convolutional layers, followed by max-pooling layers, and three fully-connected layers with a final classificatory presented with MLP (with six basic outputs, corresponding to basic emotions for emotion classification and two outputs for motion classification for typical and non-typical behavior). The input data was presented as infrared camera output.

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 11.Weka Results for Images

The dataset of the real images is also treated the same way as the dataset of the sketches. Total tree images used = 86 Total car images used = 54 Learning rate = 0.3 Momentum = 0.2 Number of epochs = 500 70% of data is used to train the neural network and the remaining 30% is used for testing the trained neural network. Figure 10 shows the neural network for images. The results are shown in Figure 11 and are as follows: Total Correct Recognition = 83.7838% Total Incorrect Recognition = 16.2162% Error Per epoch = 0.0235851

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 8. Training Neural Network for Sketches

It is described in the previous section that a mobile application is used to collect data about the sketches and the data is converted to 20x20=400 integer numbers to give it as input to Weka. Our experimentation includes only two objects for recognition i.e. trees and cars. Total tree sketches used = 175 Total car sketches used = 72 Learning rate = 0.3 Momentum = 0.2 Number of epochs = 500 70% of data is used to train the neural network and the remaining 30% is used for testing the trained neural network. Figure 8 shows the neural network for sketches. The results are shown in Figure 9 and are as follows: Total Correct Recognition = 100%

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 5. Cropped Sketch

Figure 5 shows the interface of the mobile application that is created to collect the data of the sketches. It contains a drawing canvas; where drawings are made and it also contains few text boxes which are filled with the information of the drawing. For example, if the drawing is a tree, ‘1’ is written in the tree text box. Furthermore, the undo and redo buttons help modify the drawing and the clear button clears everything drawn onto the drawing canvas.  The drawing of a tree is shown on the drawing canvas in Figure 4. It is clear from the figure that there is some empty area on top, right, left and bottom of the drawing sketch, which can cause problems while using this image for training the neural network. Therefore, there is implemented a crop function which crops the image very minutely and removes the empty space as shown in Figure 5.

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 4. Mobile Application to collect Sketch data

The drawing of a tree is shown on the drawing canvas in Figure 4. It is clear from the figure that there is some empty area on top, right, left and bottom of the drawing sketch, which can cause problems while using this image for training the neural network.

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 3. The used methodology

The training process we use in both trainings is based on neural networks, using the backpropgation algorithm. By using this algorithm, we are going to finally obtain the weights that will be used in our model to make the system recognize the input given by the user.  Once the training for the hand drawn sketches is over, we are going to get weights that will be used to recognize any inputted hand drawn sketches. For example, if the user draws a tree, the system will be able to recognize what has been drawn as a tree, using the obtained weights from the neural networks. On the other hand, once the training for the real images is over, it means that if we provide our system with a real image, for example a tree, the system will be able to recognize it using the obtained weights from the training phase.

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 6. Reducing size of Sketch

As previously described, there are a total of 1600 integers, which can be the input to the dataset for neural networks. But it is a huge number, so in order to minimize the size of inputs to the neural network the 40x40 matrix is reduced to 20x20 by skipping odd rows and columns of the original matrix. Figure 6 shows a matrix containing green and red rows and columns. If this was the 40x40 matrix, then the red part of this matrix would be skipped to convert it into a 20x20 sized matrix. Now there are only 20x20=400 values, which is a reasonable input size for the neural network.  The other part of the developed approach is to collect the data about the same two objects of the real pictures taken by the camera. These images are converted into black and white pictures and then treated the same way as the sketches, i.e. black and white pictures are also converted into integers based on the color of each pixel.

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 1. Methodology used in the research by (Egmont-Petersen et al., 2002)

Figure 1 shows the methodology used by Egmont-Petersen et al. (2002) to come up with an answer to their research question. Their study says that image recognition using neural networks goes through the stages shown in Figure 1. In our research we shall use this general proposed approach.

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 10. Training Neural Network for Images

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 2. The used approach

In this section we present in details the methodology used for this research. Figure 2 depicts the used methodology. The adopted approach is based on first training the system to be able to recognize certain sketches by providing it with various hand-drawn examples such as trees, cars and mountains. Secondly, the system has to be trained again to recognize real images by providing it with real images such as trees, cars and mountains.

BRAIN Journal-Image Finder Mobile Application Based on Neural Networks-Figure 7. Cropping image

This function takes five parameters, a WriteableBitmap type of object, the location of the starting point on x coordinate, the location of the starting point on y coordinate, width and height. It crops the image according to these parameters as shown in Figure 7.