Measuring the performance of isolated spoken Malay speech recognition using Multi-layer Neural Networks

Abstract

Abstract-This paper describes speech signal modeling techniques which are suited to high performance and robust isolated word recognition. In this study, a speech recognition system is presented, specifically an isolated spoken Malay word recognizer which uses spontaneous and formally speeches collected from Parliament of Malaysia. Currently the vocabulary is limited to 25 words that can be pronounced exactly as it written and controls the distribution of the vocalic segments. The speech segmentation task is achieved by adopted energy based parameter and zero crossing rate measure with modification to better locates the beginning and ending points of speech from the spoken words. The training and recognition processes are realized by using Multi-layer Perceptron (MLP) Neural Networks with two-layer network configurations that are trained with stochastic error back-propagation to adjust its weights and biases after presentation of every training data. The Melfrequency Cepstral Coefficients (MFCCs) has been chosen as speech extraction approach from each segmented utterance as characteristic features for the word recognizer. Recognition results showed that the performance of the two-layer networks increased as the numbers of hidden neurons increased. The best network structures average classification rate is 84.731% with (150-25) configuration. Implementation results also showed that the conjugate gradient (CG) algorithm was more accurate and reliable than the Levenberg-Marquardt (LM) algorithm for the network complexities and data size considered in this study.

Speech Recognition of Isolated Words using a New Speech Database in Sylheti

“…The authors report an overall recognition accuracy of 85% with their own database. In [8] [50]. The speech recognition models proposed in [48] for Arabic digits and words use two variants of neural network-multi-layer perceptron and Long Short-Term Memory(LSTM).…”

“…From the above discussion, it is observed that researchers are using ANN and its variants in recent times also to design ASR systems in some major languages [6], [48], [49], [50] due to their attractive characteristics as discussed in Section 1. It is to be also noted that ANN is being popularly used for digit and isolated word recognition in under-resourced languages [4], [5], [7], [8], [10]. These ANN based ASR systems are reported to deliver good recognition rates.…”

“…ii) Feature Extraction computes features from each voiced part in the pre-processed signal. Some popular features used in ASR systems are Linear Predictive Coding (LPC) coefficients [5], [6], Mel Frequency Cepstral Coefficients (MFCC) [4], [6], [7], [8], [10], short-time energy [6], i-vector [11], etc. The mel-frequency scale in MFCC coefficients being proportional to the logarithm of the linear frequency below 1 kHz, it closely reflects the human perception; and hence, MFCC features are mostly used in ASR systems.…”

Impact of acoustical voice activity detection on spontaneous filled pause classification

“…For classification purpose, each speech segment is assumed to correspond to a class which is labeled as 1 for FP and 2 for ELO. The number of input neuron will be calculated through the experiments by multiplying the cepstral order with the total frames as in (12) and (13) (12) Input Neuron No=CepstralOrder * TotalFrameNumber (13) The number of hidden neurons is determined by trial and error guided by Geometric Pyramid Rule (GPR) as in (14). The number of hidden neurons cannot be too many, otherwise, it cannot obtain good convergence rate [20].…”

“…It is a type of speech feature involving coefficients that represent audio which are derived from a type of cepstral representation of the audio clip [17]. MFCC has been successfully used in recent speech processing related work such as in non-speech detection of dysarthric speech [18], isolated spoken speech recognition [13] and continuous speech recognition [19]. A Discrete Fourier Transform (DFT) is performed on each of the windowed speech waveform with 512 DFT.…”