AN ANALYSIS OF NOISE ROBUST TECHNIQUE FOR HMM BASED PUNJABI SPEECH RECOGNITION
- Version
- Download 8
- File Size 315.53 KB
- File Count 1
- Create Date 31 July 2023
- Last Updated 31 July 2023
AN ANALYSIS OF NOISE ROBUST TECHNIQUE FOR HMM BASED PUNJABI SPEECH RECOGNITION
Ms. Mandeep kaur
Department of computer science & engineering Ghuraun ,Punjab
chandigarh university kambojmandeep30@gmail.com
Navjot Singh
Department of computer science & engineering Ghuraun ,Punjab
chandigarh university navjotsingh49900@gmail.com
Abstract— In today’s ASR engines, a set of various features extraction and modeling classifier approaches are used. Traditional front end techniques-LPCC, PNCC faces the challenges of performance degradation due to acoustic mismatch conditions. On the other end acoustic classifiers-HMM, GMM, SGMM tackle the issue of training medium vocabulary Punjabi continuous corpora. An extensive study is done to cope with these factors. Various front and back end approaches are analyzed with different baseline and hybrid methodologies. This paper tries to fulfill the gap of actual corpus performance in comparison to synthetic speech corpora. In this paper we analyze variation in acoustic mismatch and modeling information are performed using MFCC, GFCC noise robust approach at front end .
Keywords— ASR, HMM, GFCC, SGMM ,MFCC ,GMM- HMM.)
Download