LightBlog
Contact at mumbai.academics@gmail.com or 8097636691/9323040215
Responsive Ads Here

Thursday, 14 June 2018

DIFFERENT FEATURE EXTRACTION TECHNIQUES FOR AUTOMATIC SPEECH RECOGNITION: A REVIEW

ABSTRACT:-

Automatic speech recognition, which allows a usual and user-friendly communication technique among individual and device, is a dynamic research area. The speech recognition is the skill to pay attention to what we are talking about, to interpret and to perform actions based on the information spoken. This article presents a short outline of speech recognition and the various techniques like MFCC, LPC and PLP intended for feature extraction in speech recognition system. Among the three techniques i.e. MFCC, LPC, PLP, Mel frequency cepstral coefficient's (MFCC) is repeatedly used feature extraction technique in speech recognition process because it is most nearby to the real individual acoustic speech opinion.

INTRODUCTION:-

Speech is the most common type of individual communication and one of the most exciting investigation areas of the signal processing is speech processing. Speech processing is nothing but learning of language signals and the processing techniques of these signals. The signals are usually processed in a digital version, so speech processing can be viewed as a unique case of digital signal processing which is applied to speech signal. Speech Recognition is one of the plunge investigation areas in language (speech) processing, which is also known as automatic speech recognition (ASR). Speech recognition technology allows a computer to pay attention to individual voice commands and to understand individual languages. Speech recognition is the procedure of altering a given input signal into a series of words by means of an algorithm that is implemented as a computer program. That is, the speech recognition system enables a computer to recognize the words an individual speaks in a microphone or phone and convert it into readable text.

 Speech Recognition has numerous applications such as in health care, military, helicopters, telephony and other domains etc. Advancement in language (speech) technology was motivated as people wanted to develop mechanical models that allow the emulation of individual oral announcement abilities. Computers use speech processing to track voice commands and diverse individual languages. Figure 1 shows a fundamental form of speech recognition system that denote diverse phases of a scheme that contains pre-processing, speech feature extraction, classification and language model [1]. The input signal will be changed by preprocessing stage before any information can be extracted at feature extraction phase. The feature extraction phase extracts essential vectors needed for use in modeling phase after preprocessing phase. The extracted vectors have to be strong (robust) to noise for improved accuracy. The language text is recognized by classification phase by using extracted features and a language template where the language template contains syntax and semantics related to the responsible language which help the classifiers to identify the input statement

No comments:

Post a Comment