Investigation on speech recognition Accuracy via Sphinx toolkits

Automatic speech recognition (ASR) is the process that concerns the permit of the machines to accept speech input from humans and interpret it with the highest precision probability. The availability of usable open-source software playing a remarkable role in the popularization of this objective. The aim of this paper is to evaluate the Sphinx open-source speech recognition toolkit. Specifically, the decoders Pocketsphinx and Sphinx-4 are compared in terms of usability and expense of recognition accuracy. The evaluation presented in this paper was carried out using Amazigh language. Our system was trained to recognize the ten first Amazigh digits by using 5 Hidden Markov model states (HMMs) with 16 Gaussian mixture models (GMMs) and the Mel frequency spectral coefficients (MFCCs). Our obtained experimental results show that the best recognition rate was recorded by Pocketsphinx toolkit.