Continuous Speech Recognition of Punjabi Language

Abstract

Speech is one mode of communication. It is a way by which we can share thoughts, emotions, and data between human beings. The speech recognition is a technique to change an acoustic signal captured by an input device, into the sequence of words. Speech recognition technique is an interesting area to work on, where main result is yet to achieve i.e., a system which can recognize spoken discourse by speakers in almost every condition. Work on speech recognition is not new, for many years people have been trying to make our machines hear, understand, and also speak our natural language. But the issue newlineis to implement the robust speech recognition system which works efficiently newlineindependent of environment as well as speaker. newlineThe earliest attempts made for automatic speech recognition systems through exploiting a fundamental idea of acoustic phonetics. But these systems were not successful because of computational cost, requirement of phonetics knowledge, and poor recognition performance. And these are due to issues in reliably extracting phonetic data from the speech signal. After that, pattern matching approaches become popular. In these techniques, no phonetic based knowledge is required. The features are extracted from speech signal using feature extraction techniques. Some of the features are trained, and newlinethe other features are used for testing using classification techniques. Today advanced newlineASR systems for pattern matching are based on Hidden Markov Model. As HMM ASR newlinesystems were insensitive by sound class with fewer computations and high recognition accuracy. newlineResearch activities has been done on different languages based automatic speech recognition system. Research work has been reported in area of speech recognition for the languages like English, Arabic and Chinese etc. In recent years work has also been reported in the literature in Hindi language.