Tran Huy Dat

Sub-band Common Spatial Pattern (SBCSP) for Brain-Computer Interface

Brain-computer interface (BCI) is a system to translate humans thoughts into commands. For electroencephalography (EEG) based BCI, motor imagery is considered as one of the most effective ways. Different imagery activities can be classified based on the changes in mu and/or beta rhythms and their spatial distributions. However, the change in these rhythmic patterns varies from one subject to another. This causes an unavoidable time-consuming fine-tuning process in building a BCI for every subject. To address this issue, we propose a new method called sub-band common spatial pattern (SBCSP) to solve the problem. First, we decompose the EEG signals into sub-bands using a filter bank. Subsequently, we apply a discriminative analysis to extract SBCSP features. The SBCSP features are then fed into linear discriminant analyzers (LDA) to obtain scores which reflect the classification capability of each frequency band. Finally, the scores are fused to make decision. We evaluate two fusion methods: recursive band elimination (RBE) and meta-classifier (MC). We assess our approaches on a standard database from BCI Competition III. We also compare our method with two other approaches that address the same issue. The results show that our method outperforms the other two approaches and achieves similar result as compared to the best one in the literature which was obtained by a time-consuming fine-tuning process.

Heart sound as a biometric

In this paper, we propose a novel biometric method based on heart sound signals. The biometric system comprises an electronic stethoscope, a computer equipped with a sound card and the software application. Our approach consists of a robust feature extraction scheme which is based on cepstral analysis with a specified configuration, combined with Gaussian mixture modeling. Experiments have been conducted to determine the relationship between various parameters in our proposed scheme. It has been demonstrated that heart sounds should be processed within segments of 0.5s and using the full resolution in frequency domain. Also, higher order cepstral coefficients that carry information on the excitation proved to be useful. A preliminary test of 128 heart sounds from 128 participants was collected to evaluate the uniqueness of the heart sounds. The HTK toolkit produces a 99% recognition rate with only one mismatch. Next, a more comprehensive test consisting almost 1000 heart sounds collected from 10 individuals over a period of 2 months yields a promising matching accuracy of 96% using the proposed feature and classification algorithm. A real-time heart sound authentication system is then built and can be used in two modes: to identify a particular individual or to verify an individuals claimed identity.

Generalized gamma modeling of speech and its online estimation for speech enhancement

Generalized gamma modeling and its online method of parameter estimation of speech spectral magnitude are proposed for MAP based speech enhancement systems. Generalized gamma modeling is shown to be a natural extension of the Gaussian modeling of speech spectral component distribution, and is therefore, able to fit the prior distribution better than the conventional method. An online parameter estimation method for the gamma distribution, based on a moment matching method, is then proposed. The effectiveness of the proposed methods are confirmed by improvement in both SNR and ASR using the AURORA2 standard database, where about 4 dB improvement in SNR and 20% improvement in relative ASR performance are obtained.

Feature Selection Based on Fisher Ratio and Mutual Information Analyses for Robust Brain Computer Interface

This paper proposes a novel feature selection method based on two-stage analysis of Fisher ratio and mutual information for robust brain computer interface. This method decomposes multichannel brain signals into subbands. The spatial filtering and feature extraction is then processed in each subband. The two-stage analysis of Fisher ratio and mutual information is carried out in the feature domain to reject the noisy feature indexes and select the most informative combination from the remaining. In the approach, we develop two practical solutions, avoiding the difficulties of using high dimensional mutual information in the application, that are the feature indexes clustering using cross mutual information and the latter estimation based on conditional empirical PDF. We test the proposed feature selection method on two BCI data sets and the results are at least comparable to the best results in the literature. The main advantage of proposed method is that the method is free from any time-consuming parameter tweaking and therefore suitable for the BCI system design.

Electrocorticographic signal classification based on time-frequency decomposition and nonparametric statistical modeling

In this paper, we propose a novel statistical framework based on time-frequency decomposition and nonparametric modelling of electrocortical (ECoG) signals in the context of a Brain Computer Interface. The proposed method decomposes the ECoG signals into subbands (with no down-sampling) using Gabor filters. The subband signals are then encoded using a nonparametric statistical modeling and the distance between the resulting empirical distributions is as used as the classification criterion. Cross-validation experiments were carried out to pre-select the channel (from the multi-channel sources) and subbands which can archive the best classification scores. The proposed framework has been evaluated using Data Set I from the BCI Competition III and results indicate a superiority over conventional vector quantization method particularly when the number of training samples is small. It was found that the proposed nonparametric distribution modeling based on empirical inverse cumulative distribution distance is fast, robust and applicable to the mobile systems

Introduction to NeuroComm: a Platform for Developing Real-Time EEG-based Brain-Computer Interface Applications

NeuroComm is a platform to develop real time brain computer interface (BCI) applications. This paper introduces the basic modules of this platform and discusses some implementation issues. With a user management module, our system is user friendly and suitable for multiple users. Also, with flexible configuration files and signal processing algorithm libraries, it is easier to integrate multiple BCI applications into one system. The NeuroComm platform also acts as a flexible tool for BCI research.

Single and multi-channel approaches for distant speech recognition under noisy reverberant conditions: I2R'S system description for the ASpIRE challenge

In this paper, we introduce the system developed at the Institute for Infocomm Research (I2 R) for the ASpIRE (Automatic Speech recognition In Reverberant Environments) challenge. The main components of the system are a front-end processing system consisting of a distributed beam-forming algorithm, that performs adaptive weighting and channel elimination, a speech dereverberation approach using a maximum-kurtosis criteria, and a robust voice activity detection (VAD) module based on using the sub-harmonic ratio (SHR). The acoustic back-end consists of a multi-conditional Deep Neural Network (DNN) model that uses speaker adapted features combined with a decoding strategy that performs semi-supervised DNN model adaptation using weighted labels generated by the first-pass decoding output. On the single-microphone evaluation, our system achieved a word error rate (WER) of 44.8%. With the incorporation of beamforming on the multi-microphone evaluation, our system achieved an improvement in WER of over 6% to give the best evaluation result of 38.5%.

Affective social interaction with CuDDler robot

This paper introduces an implemented affective social robot, called CuDDler. The goal of this research is to explore and demonstrate the utility of a robot that is capable of recognising and responding to a users emotional acts (i.e., affective stimuli), thereby improving the social interactions. CuDDler uses two main modalities; a) audio (i.e., linguistics and non-linguistics sounds) and b) visual (i.e., facial expressions) to recognise the users emotional acts. Similarly, CuDDler has two modalities; a) gesture and b) sound to respond or express its emotional responses. During the TechFest 2012 event, CuDDler successfully demonstrated its capability of recognising the users emotional acts and responding its expression accordingly. Although, CuDDler is still in its early prototyping stage, the preliminary survey results indicate that the CuDDler has potential to not only aid in human-robot interaction but also contribute towards the long term goal of multi-model emotion recognition and socially interactive robot.

Multichannel Speech Enhancement Based on Speech Spectral Magnitude Estimation Using Generalized Gamma Prior Distribution

We present multichannel speech enhancement method based on MAP speech spectral magnitude estimation using a generalized gamma model of speech prior distribution, where the model parameters are adapted from actual noisy speech in a frame-by-frame manner. The utilization of a more general prior distribution with its online estimation is shown to be effective for speech spectral estimation. We tested the proposed algorithm in an in-car speech database and obtained significant improvements on the speech recognition performance, particularly under nonstationary noise conditions such as music, air-conditioner and open window

Maximum a posterior probability and cumulative distribution function equalization methods for speech spectral estimation with application in noise suppression filtering

In this work, we develop and compare noise suppression filtering systems based on maximum a posterior probability (MAP) and cumulative distribution function equalization (CDFE) estimation of speech spectrum. In these systems, we use a double-gamma modeling for both the speech and noise spectral components, in which the distributions are adapted to the actual parameters in each frequency bin. The performances of the proposed systems are tested using the Aurora database they are shown to be better than conventional systems derived from the MMSE method. Whereas the MAP-based method performed best in the SNR improvement, the CDFE-based system provides a lower musical noise level and shows a higher recognition rate.