Training and adapting MLP features for Arabic speech recognition

Abstract

Features derived from Multi-Layer Perceptrons (MLPs) are becoming increasingly popular for speech recognition. This paper describes various schemes for applying these features to state-of-the-art Arabic speech recognition: the use of MLP-features for short-vowel modelling in graphemic systems; rapid discriminative model training by standard PLP feature lattice re-use; and MLP feature adaptation using Linear Input Networks (LIN). The use of rapid training using MLP features and their use for short-vowel modelling and LIN adaptation gave reductions in word error rate. However significant improvements over explicit short-vowel modelling with standard multi-pass adaptation were not obtained, although they were useful in combination.

Analysis of MLP-Based Hierarchical Phoneme Posterior Probability Estimator

“…Any two points in the plot correspond to systems with the same number of parameters, and can be calculated using. 5 It can be seen from the figure that:…”

“…As the temporal context on the posterior features at the second MLP is increased, the total number of parameters in the MLP is kept constant by appropriately reducing the size of its hidden layer. 5 In the case of single MLP estimator, as the temporal context on the acoustic features is increased, the total number of parameters is kept constant, and equal to those in the hierarchical system (sum of the parameters in both the MLPs). Any two points in the plot correspond to systems with the same number of parameters, and can be calculated using.…”

“…. y N (t)] ′ system characterized by (5). This system can be viewed as N parallel, multi-input, single-output, nonlinear, time-invariant systems, and represented by…”

“…Ultilayer perceptron (MLP) classifier based acoustic modeling is being extensively used in state-of-the-art automatic speech recognition (ASR) systems [1][2] [3][4] [5]. The MLP is typically trained using standard acoustic features such as mel frequency cepstral coefficients or perceptual linear predictive coefficients with a certain temporal context.…”

Cross-Word Arabic Pronunciation Variation Modeling Using Part of Speech Tagging

“…(Mokhtar & El-Abddin, 1996) represented the techniques and algorithms used to model the acoustic-phonetic structure of Arabic speech recognition using HMMs. (Park et al , 2009) explored the training and adaptation of multilayer perceptron (MLP) features in Arabic ASRs. They used MLP features to incorporate short-vowel information into the graphemic system.…”