Hadri Hussain

Speaker Verification Using Gaussian Mixture Model (GMM)

This paper applies GMM for SV on Malay speech. The speaker models were trained on maximum likelihood estimated. The system was evaluated with 23 client speakers with 56 imposters. Malay clean speech data was used. 20 training of 3.5s utterances are used. The best performance achieved using 256-Gaussian imposter model and 32-Gaussian client model gave 3.01% of EER.

Analysis of ECG biosignal recognition for client identifiction

The most common application for a recognition system of speech signal, finger print, iris, etc. are used for biometrie applications. While other biometric signals like electrocardiogram (ECG) and the Heart Sound (HS) are generally used to identify cluster-related diseases. Nonetheless, performance of a traditional biometric system can be easily compromised as it is prone to spoof attack. This paper proposes a unimodal biometric security system that is based on ECG. Physiological biometrics characteristic are based on a human bodys, such as the hand geometry, face, palm, ECG and even brain signal. The biosignal data collected by a biometric system would initially be segmented. The Mel-Frequency Cepstral Coefficients (MFCC) method is used for extracting each segmented feature. The Hidden Markov Model (HMM) is used to model the client, and categorize unknown input based on the model. The recognition system involved training and testing of the collected features, known as Client Identification (CID). In this paper, 20 clients were tested with this developed system. The best overall performance for 20 clients at 16 kHz was 71.4% for ECG trained at 50% of the training data, while the worst overall performance was 66.6% for 30% training data.

Biosignal recognition for patients identification

A biometric security system has becoming an important application in client identification and verification system. A conventional biometric system is normally based on unimodal biometric that depends on either behavioral or physiological information for authentication purposes. The behavioral biometric depends on human body biometric signal (such as speech) and biosignal biometric (such as electrocardiogram and phonocardiogram or heart sound). The speech signal is commonly used in a recognition system in biometric, while the electrocardiogram and the heart sound have been used to identify a persons diseases, uniquely related to its cluster. However, the conventional biometric system is liable to spoof attack, which affect the performance of the system. In this paper, a multimodal biometric security system is developed, which is based on biometric signal of electrocardiogram and heart sound. The biosignal data involved in the biometric system initially segmented, with each segment Mel Frequency Cepstral Coeffiecients method is exploited for extracting the features. The Hidden Markov Model is used to model the client and to classify the unknown input with respect to the modal. The recognition system involved training and testing session that is known as Client Identification. In this project, twenty clients are tested with the developed system. The best overall performance for 20 clients at 44 kHz was 93.92% for electrocardiogram train at 70% of the training data however the worst overall performance was also electrocardiogram at an increment of data client of 63 clients at 79.91% for 30% training data. It can be concluded that the difference multimodal biometric has a substantial effect on performance of the biometric system and with the increment of data, even with higher sampling rate, the performance still decreased slightly as predicted.