Bangla User Adaptive Word Speech Recognition: Approaches and Comparisons

The paper presents Bangla word speech recognition using two novel approaches with a comprehensive analysis. The first approach is based on spectral analysis and fuzzy logic and the second one uses Mel-Frequency Cepstral Coefficients (MFCC) analysis and feed-forward back-propagation neural networks. As human speech is imprecise and ambiguous, fuzzy logic – the base of which is indeed linguistic ambiguity, could serve as a precise tool for analyzing and recognizing human speech. The authors’ systems revolve around the visual representations of voiced signals – the Fourier energy spectrum and the MFCC. The essences of a Fourier energy spectrum and the MFCC are matrices that include information about properties of a sound by storing energy and frequency in discrete time. The decision making process of their systems is based on fuzzy logic and neural networks. Experimental results demonstrate that their fuzzy logic based system is 86% accurate whereas the Artificial Neural Networks (ANN) based system is 90% accurate compared to a commercial Hidden Markov Model (HMM) based speech recognizer that shows 73% accuracy on an average. Moreover, the authors’ research derives that, even though ANN gives a better recognition accuracy than the fuzzy logic based system, the fuzzy logic based system is more accurate when it comes to “more difficult” or “polysyllabic” words. In terms of runtime performance, the fuzzy logic based system outperforms the ANN based Bangla speech recognition system.