Setha Pan-ngum

EEG-Based Emotion Recognition Using Deep Learning Network with Principal Component Based Covariate Shift Adaptation

Automatic emotion recognition is one of the most challenging tasks. To detect emotion from nonstationary EEG signals, a sophisticated learning algorithm that can represent high-level abstraction is required. This study proposes the utilization of a deep learning network (DLN) to discover unknown feature correlation between input signals that is crucial for the learning task. The DLN is implemented with a stacked autoencoder (SAE) using hierarchical feature learning approach. Input features of the network are power spectral densities of 32-channel EEG signals from 32 subjects. To alleviate overfitting problem, principal component analysis (PCA) is applied to extract the most important components of initial input features. Furthermore, covariate shift adaptation of the principal components is implemented to minimize the nonstationary effect of EEG signals. Experimental results show that the DLN is capable of classifying three different levels of valence and arousal with accuracy of 49.52% and 46.03%, respectively. Principal component based covariate shift adaptation enhances the respective classification accuracy by 5.55% and 6.53%. Moreover, DLN provides better performance compared to SVM and naive Bayes classifiers.

Real-time EEG-based happiness detection system.

We propose to use real-time EEG signal to classify happy and unhappy emotions elicited by pictures and classical music. We use PSD as a feature and SVM as a classifier. The average accuracies of subject-dependent model and subject-independent model are approximately 75.62% and 65.12%, respectively. Considering each pair of channels, temporal pair of channels (T7 and T8) gives a better result than the other area. Considering different frequency bands, high-frequency bands (Beta and Gamma) give a better result than low-frequency bands. Considering different time durations for emotion elicitation, that result from 30 seconds does not have significant difference compared with the result from 60 seconds. From all of these results, we implement real-time EEG-based happiness detection system using only one pair of channels. Furthermore, we develop games based on the happiness detection system to help user recognize and control the happiness.

Emotion classification using minimal EEG channels and frequency bands

In this research we propose to use EEG signal to classify two emotions (i.e., positive and negative) elicited by pictures. With power spectrum features, the accuracy rate of SVM classifier is about 85.41%. Considering each pair of channels and different frequency bands, it shows that frontal pairs of channels give a better result than the other area and high frequency bands give a better result than low frequency bands. Furthermore, we can reduce number of pairs of channels from 7 to 5 with almost the same accuracy and can cut low frequency bands in order to save computation time. All of these are beneficial to the development of emotion classification system using minimal EEG channels in real-time.

Electronic stethoscope prototype with adaptive noise cancellation

Electronic stethoscope is developed to improve the limitations and add functions to a conventional stethoscope. In electronic stethoscope, sound is transformed into electrical signal which offers various benefits including signal quality improvement, signal record and signal transfer via wireless. Electronic stethoscope in the market is expensive. Therefore, this research aims to design and develop a low-cost electronic stethoscope prototype that can filter frequency, reduce external noise and make real-time digital signal processing.

A study of low-cost, robust Assistive Listening System (ALS) based on digital wireless technology

Purpose. We have developed a simple, low-cost digital wireless broadcasting system prototype, intended for a classroom of hearing impaired students. The system is designed to be a low-cost alternative to an existing FM system. Method. The system implemented is for short-range communication, with a one-transmitter, multiple-receiver configuration, which is typical for these classrooms. The data is source-coded for voice-band quality, FSK modulated, and broadcasted via a 915 MHz radio frequency. A DES encryption can optionally be added for better information security. Results. Test results show that the system operating range is approximately ten metres, and the sound quality is close to telephone quality as intended. We also discuss performance issues such as sound, power and size, as well as transmission protocols. Conclusions. The test results are the proof of concept that the prototype is a viable alternative to an existing FM system. Improvements can be made to the systems sound quality via techniques such as channel coding, which is also discussed.

Mean travel speed estimation using GPS data without ID number on inner city road

Due to the expansive use of GPS, GPS data can be used to provide valuable travel time and the travel speed data for the traffic information system. However, to access ID number for personal car would have problems with privacy. Thus, mean travel speed (MTS), which requires individual vehicle tracking, cannot be calculated directly. In this research, we try to estimate the MTS of the road by using GPS data without ID number. The estimation is done by considering the relationship between MTS and time mean speed (TMS), and experiment with relation between MTS and estimated space mean speed (eSMS). In addition, vehicle speed has high variance on inner city roads. Therefore, we proposed grouping method of spot speed data on the road segment, which is called segmentation, to reduce the traffic variance and analyze the collected data for MTS estimation. The results show the correlation between TMS and MTS is 0.94 and the relationship graphs between TMS and MTS have a linear trend line. Hence, TMS and MTS are very correlative. In summary, MTS estimation can be improved and developed into the model or equation if TMS is calculated under short segment (50 m), low traffic variance data, and under suitable time period (5 minutes).

Brain signal detection methodology for attention training using minimal EEG channels

This paper presents the methodology to find the appropriate locations on the scalp for detecting EEG signal for attention training. The aim was to use low cost commercial EEG sensing device to select four positions which provide strong attention related signals. We aim to use the results of this work to develop a low cost attention training set. Three experiments were conducted to collect raw EEG data from 10 examiners. All 3 experiments made the examiners focus on the specific task to stimulate attention. First experiment is to read a light-content book, second is to figure out a hidden 3D image and the last one is to answer a general questionnaire on a PC. As the result, the suggested locations for detecting Alpha wave are AF3 and F7. The suggested locations for detecting Beta wave are FC6 and F8. It also shows that Alpha wave in the relaxation state higher than Alpha wave in the attention state and Beta wave in the attention state higher than Beta wave in the relaxation state.

A study of low-cost, robust assistive listening system based on UHF wireless technology

In this paper, we describe the development of a low cost wireless UHF broadcasting system prototype. It is designed for a short range application, with up to ten receivers, such as one used in a classroom for hearing impaired students. This system is intended to be a low cost alternative to an existing FM system, with additional capability in security with an encryption function. Additionally a multimode function including one-to-one, one-to-many and broadcast is added. Test results show that the system operating range is approximately ten metres, and the sound quality is close to telephone quality. The broadcasting mode and system range fit the requirements for small classroom use.

Development of a Low Cost Assistive Listening System for Hearing-Impaired Student Classroom

This paper describes the design, development, and tests of a low cost ALS. It was designed for hearing-impaired student classrooms. It utilised digital wireless technology and was aimed to be an alternative to a popular FM ALS. Key specifications include transmitting in 2.4 GHz ISM band with eight selectable transmission channels, battery operated and chargeable, pocket size, and ranged up to thirty metres. Audio characteristics and user tests show that it is comparable to a commercial system, currently employed in our partner school. The results also show that wearing an ALS clearly improves hearing of hearing-impaired students. Long-term usage by school children will be monitored to evaluate the system robustness and durability.

Development of the inner city following-lane changing model and meandering model of motorcycles

As in many Asian cities, Bangkok motorcycle riders tend to ride along a traffic lane with a car or other motorcycles. They also meander through spaces between vehicles at intersections. These behaviors should have impact on the traffic throughputs and vehicle queuing times. Since motorcycles accounts for 20% of all Bangkok vehicles, this impact should be sizable. We, therefore, developed an inner city motorcycle behavior model, based on the city riders. In the following-lane changing model, a motorcycle could share a traffic lane with another motorcycle or car. At traffic lights, meandering model would simulate motorcycles slipping through spaces between vehicle queues. The model was developed on Aimsun NG. Real traffic data used to develop the model was obtained from observations and video-recording. The recorded videos were analyzed to get a probability of each motorcycle lane changing pattern and a probability of each motorcycle meandering pattern. Simulation results indicate that the errors in travel time and speed are within 8 percents, and the error in time in queue is within 8 percents of the test data.