M. A. Yusnita

Malaysian English accents identification using LPC and formant analysis

In Malaysia, most people speak several varieties of English known as Malaysian English (MalE) and there is no uniform version because of the existence of multi-ethnic population. It is a common scenario that Malaysians speak a particular local Malay, Chinese or Indian English accent. As most commercial speech recognizers have been developed using a standard English language, it is a challenging task for achieving highly efficient performance when other accented speech are presented to this system. Accent identification (AccID) can be one of the subsystem in speaker independent automatic speech recognition (SI-ASR) system so that deterioration issue in its performance can be tackled. In this paper, the most important speech features of three ethnic groups of MalE speakers are extracted using Linear Predictive Coding (LPC), formant and log energy feature vectors. In the subsequent stage, the accent identity of a speaker is predicted using K-Nearest Neighbors (KNN) classifier based on the extracted information. Prior, the preprocessing parameters and LPC order are investigated to properly extract the speech features. This study is conducted on a small set speech corpus developed as pilot study to determine the feasibility of automatic AccID of MalE speakers which has never been reported before. The experimental results indicate a highly promising recognition accuracy of 94.2% upon feature fusion sets of LPC, formants and log energy.

Analysis of EEG signals during relaxation and mental stress condition using AR modeling techniques

Electroencephalography (EEG) is the most important tool to study the brain behavior. This paper presents an integrated system for detecting brain changes during relax and mental stress condition. In most studies, which use quantitative EEG analysis, the properties of measured EEG are computed by applying power spectral density (PSD) estimation for selected representative EEG samples. The sample for which the PSD is calculated is assumed to be stationary. This work deals with a comparative study of the PSD obtained from resting and mental stress condition of EEG signals. The power density spectra were calculated using fast Fourier transform (FFT) by Welchs method, auto regressive (AR) method by Yule-Walker and Burgs method. Finally a neural network classifier used to classify these two conditions. It is found that maximum classification accuracy of 91.17% was obtained for the Burg Method compared to Yule Walker and Welch Method technique.

Classification of speaker accent using hybrid DWT-LPC features and K-nearest neighbors in ethnically diverse Malaysian English

Accent is a major cause of variability in automatic speaker-independent speech recognition systems. Under certain circumstances, this event introduces unsatisfactory performance of the systems. In order to circumvent this deficiency, accent analyzer in preceding stage could be a smart solution. This paper proposes a rather new approach of hybrid way to optimize the extraction of accent from speech utterances over other facets using linear predictive coefficients (LPC) derived from discrete wavelet transform (DWT). The constructed features were used to model an accent recognizer, implemented based on K-nearest neighbors. Experimental results showed that the hybrid dyadic-X DWT-LPC features were highly correlated to the Malay, Chinese and Indian accents of Malaysian English speakers through an increase of classification rate of 9.28% over the conventional LPC method.

Phoneme-based or isolated-word modeling speech recognition system? An overview

In this paper speech theories and some methodological concerns about feature extraction and classification techniques widely used in speech recognition system are surveyed and discussed. The shortage of isolated word speech recognition is addressed as compared to its phoneme-based counterpart. This paper could be regarded as a very early stage towards methodology establishment in searching for better accuracy and less complexity system which has more generic properties. It is hoped that the system can classify speech regardless of the varieties across languages or accents. Speaker independency (SI) manner speech recognition system is required for this application and in fact, in many other potential applications as much as a telephonic network (large database consists of many different speakers) is a primary requirement. Isolated-word ASR for fixed vocabularies has been successfully implemented using HMM, ANN and SVM but suffers from lack of adaptability to other languages and increase in complexity as number of vocabularies increases. Conversely, phonemes, the smallest unit of human speech sounds are apparently more feasible to represent the basic building block for cross-language mapping. In fact, the phonetic transcription systems such as IPA and SAMPA are widely recognized and standardized for several languages in the world. This paper intends to investigate the phoneme-based potential as language independent phonetic units to overcome the lack of available training data so as to achieve a more generic speech recognizer.

Feature space reduction in ethnically diverse Malaysian English accents classification

In this paper we propose a reduced dimensional space of statistical descriptors of mel-bands spectral energy (MBSE) vectors for accent classification of Malaysian English (MalE) speakers caused by diverse ethnics. Principle component analysis (PCA) with eigenvector decomposition approach was utilized to project this high-dimensional dataset into uncorrelated space through the interesting covariance structure of a set of variables. This delimitates the size of feature vector necessary for good classification task once significant coordinate system is revealed. The objectives of this paper have three-fold. Firstly, to generate reduced size feature vector in order to decrease the memory requirement and the computational time. Secondly, to improve the classification accuracy. Thirdly, to replace the state-of-the-art mel-frequency cepstral coefficients (MFCC) method that is more susceptible to noisy environment. The system was designed using K-nearest neighbors algorithm and evaluated on 20% independent test dataset. The proposed PCA-transformed mel-bands spectral energy (PCA-MBSE) on MalE database has proven to be more efficient in terms of space and robust over the baselines MFCC and MBSE. PCA-MBSE achieved the same accuracy as the original MBSE at 66.67% reduced feature vector and tested to be superiorly robust under various noisy conditions with only 10.48% drop in the performance as compared to 16.81% and 48.01% using MBSE and MFCC respectively.

On the Use of Spectral Feature Fusions for Enhanced Performance of Malaysian English Accents Classification

Accent problem is a current issue that degrades the intelligibility and performance of speech recognition (ASR) systems. Despite English accents have been extensively researched in the United States, Britain, Australia, China, India, and Singapore, the study of Malaysian English (MalE) is still at infancy. There is till date, very limited evidence to corroborate how ethnically diverse accents in MalE of its three main ethnics can be identified from their speech signals. Most studies about MalE tackles issues from the view point of attitudinal studies and making use of human perceptual analysis. Instead, this paper presents experimental methods by means of acoustical analysis and machine learning techniques. In order to enhance the performance of accent classifier to classify the Malay, Chinese, and Indian accents this paper proposes fusion techniques of popularly known mel-frequency cepstral coefficients (MFCC) and linear prediction coefficients (LPC) with formants termed here as spectral feature fusions (SFFs). In these SFFs feature extractors, the main spectral features are fused with five usable formants and the extracted features are used to model K-nearest neighbors and artificial neural networks (ANN). Using independent test samples technique, gender-dependent accent classifiers were evaluated. Experimental results showed that the proposed SFFs surpassed the baseline features by 7.8 and 3.9 % increment of the classification rates for the LPC-formants and MFCC-formants fusions, respectively. The highest accuracies yielded for the fusion of MFCC and formants were 96.4 and 92.5 % on the male and female datasets. Speaking of LPC-formants fusion, the results were also promising, i.e., 92.6 and 88.8 % on the male and female datasets, respectively.

Robustness Analysis of Feature Extractors for Ethnic Identification of Malaysian English Accents Database

Accent is fascinating human speech behavior that can be used to mark personal identity and social characteristics of its bearer. However, it also has potential to bias social interaction which includes prestige, job competency and ethnic discrimination. Albeit many successful methods have been deployed in the past to identify a speaker accent, the success rates are most likely database-dependent. This study aims to inquire about identification of Malaysian English (MalE) accents caused by ethnic diversities in this country. Robustness analysis was conducted using seven noisiness levels by corrupting the speech signals with additive white Gaussian noise (AWGN) to investigate the performance of four different schemes of feature extractors under clean and noisy conditions. These methods are filter bank analysis consists of mel-frequency cepstral coefficients (MFCC) and a new set of formulated features named as descriptors of mel-bands spectral energy (MBSE). Principle component analysis (PCA) was utilized to transform to another new features called PCA-MBSE. Second, vocal tract analysis consists of linear prediction coefficients (LPC) and formant frequencies (formants). Third, hybrid analysis consists of discrete wavelet transform (DWT) and LPC. The last scheme is fusions of spectral features (SFFs) of MFCC with formants and LPC with formants. Experimental results showed that SFFs techniques possess more sturdy noise resistivity than MFCC, LPC, MBSE, and DWT-derived LPC features. Similarly, PCA-transformed MBSE was just moderately affected as compared to the original features. While PCA-MBSE only caused a performance drop of 15 % in average and the SFFs were just slightly affected by the AWGN from 8 to 13 % drop, the percentage drop of other feature sets were fairly above 30 %.

Multi-resolution Analysis of Linear Prediction Coefficients using Discrete Wavelet Transform for Automatic Accent Recognition of Diverse Ethnics in Malaysian English

Accent is a major cause of variability in speaker-independent automatic speech recognition (ASR) systems. Under certain circumstances, this behavioral factor introduces unsatisfactory performance of the systems. Thus, accent analyzer in the preceding stage of the ASR system becomes a promising solution. This paper proposes a multi-resolution approach which applies discrete wavelet transform (DWT) to conventional linear prediction coefficients (LPC) to optimize the extraction of accent from speech utterances in Malaysian English. This paper introduces a multi-numbered LPC (dyadic DWT-LPC) using a defined scale named as level dyadic division scale and an equal-numbered LPC (uniform DWT-LPC) approaches. Using the extracted features, accent models based on K-nearest neighbors were developed. Experimental results showed that the proposed multi-resolution dyadic DWT-LPC and uniform DWT-LPC features surpassed the conventional LPC by significant increases of classification rate of 12.7 and 17.0 % respectively. The promising results of 93.4 % and 88.5 % were achieved using the proposed methods.

Statistical formant descriptors with linear predictive coefficients for accent classification

Accent is a special trait of human speech that can deliver some information about a speakers background. At the same time it is one of the profound factors that affects the intelligibility and performance of speech recognition systems (ASRs) if not delicately handled. Normally accent recognizer in the preceding stage offers subsystem training or adaptation strategy to improve the ASRs. Formant analysis is one of the effective techniques used to extract accent information in speech. In this paper we propose a novel way of modifying formants using statistical descriptors and fusion with linear predictive coefficients (LPC). As a result, the deviation of scores from the means can be reduced and resulted in better accuracy rate. This work was based on database of accents in Malaysian English that are ethnically diverse in nature. Experimental results showed that the proposed fusion of LPC with statistically derived fmntRRS has achieved an increase of 7.61% in the accuracy rate over using LPC alone in the quest to classify three-accent problem.