Speech Recognition Techniques: A Comparative Review

Speech Recognition Techniques: A Comparative Review

Mr. Vasu Jeevan Naik

Masters of Engineering,

Department of Information Technology, Goa College of Engineering, Ponda, Goa 403401

vasunaik1999@gmail.com

Dr. Nilesh B. Fal Dessai

Head of the Department,

Department of Information Technology, Goa College of Engineering, Ponda, Goa 403401

nfd@gec.ac.in

Keywords: Speech Recognition, Speech to Text, Hidden Markov Models, Recurrent Neural Networks, Connectionist Temporal Classification, Convolutional Neural Networks.

Abstract— Speech recognition, also known as speech-to-text or automatic speech recognition, is a technology that enables the transformation of spoken words into written or machine-readable text. This technology has numerous applications such as speech-to-text dictation, voice command, voice-enabled interfaces, automated call centers, hand-free computing, and translation. The working of a speech recognition system involves several steps such as data collection, data preprocessing, feature extraction, model selection, model training, model evaluation, and model deployment. There are various algorithms/methods used to build speech recognition systems, including Hidden Markov Models (HMMs), Recurrent Neural Networks (RNNs), Connectionist Temporal Classification (CTC), Gaussian Mixture Models (GMM), Deep Neural Networks (DNN) and Convolutional Neural Networks (CNN).  In this paper, we present an overview of speech recognition and four methods: Hidden Markov Models (HMMs), Recurrent Neural Networks (RNNs), Connectionist Temporal Classification (CTC), and Convolutional Neural Networks (CNN) used for building speech recognition system. Also, we have discussed how these methods can be evaluated using several metrics such as Word Error Rate (WER), Word Accuracy, Character Error Rate (CER), Precision, Recall, F1 Score, Perplexity, Latency, and Speaker Dependence/Independence.