Open AccessFeature Extraction and Processing Analysis in Speech Recognition
Author(s) -
Mrinal Paliwal,
Pankaj Saraswat
Publication year - 2021
Publication title -
international journal of innovative research in computer science and technology
Language(s) - English
Resource type - Journals
ISSN - 2347-5552
DOI - 10.55524/ijircst.2021.9.6.12
Subject(s) - speech recognition , computer science , speech processing , feature (linguistics) , distortion (music) , speaker recognition , feature extraction , linear predictive coding , voice activity detection , variety (cybernetics) , signal (programming language) , identification (biology) , stress (linguistics) , speech technology , voice analysis , artificial intelligence , linguistics , amplifier , computer network , philosophy , botany , bandwidth (computing) , biology , programming language
The difficulties with automated identification and synthesis of various speech patterns have become significant research issues in recent years. Stressinduced speech characteristics were compared to normal speech in a feature analysis. Due to stress, the performance of Stressed speech recognition decreases substantially. In the speech communication system, the voice signal is transmitted, stored, and processed in a variety of ways. The speech signal must be delivered in such a way that the information content may be easily extracted from human listeners or machine automation. To enhance speech recognition performance, a stressed compensation method is employed to compensate for stress distortion. To identify different moods in speech signals, these features are collected and assessed in English. The variations in glottal excitement of common speaking patterns are examined in depth in this article. The sinusoidal model effectively describes the different stress classes in a speech signal, according to the results. When it comes to detecting emotions in a pressured speaker, sinusoidal features outperform linear prediction features.