Please use this identifier to cite or link to this item:
Title: Characterization of noise associated with forensic speech samples for speech recognition and speaker identification
Researcher: Jiju, P V
Guide(s): Sharma, R M
Singh, C P
Keywords: Forensic Science
Fourier s Theorem
Convolution Theorem
Upload Date: 18-Jul-2013
University: Punjabi University
Completed Date: 2009
Abstract: Speech is the most extensively used mode of communication among human. If recorded properly it is also counted as one of the prime physical evidences by the court of law all over the world whether it is a call of extortion, threatening, bribery, kidnapping or an act of terrorism. Noisy or corrupt audio recordings loose its value as evidence if the Judge and Jury cannot understand the text of conversation which demands noise-corrupted speech to be made recognizable and hence the process of noise cancellation/reduction is necessary in such cases for Speech Recognition and Speaker Identification. The task of noise reduction in Forensic case examination is in two folds, either restricted filtering for Speaker Identification or Intelligibility Enhancement (Speech Recognition) by noise subtraction for Jury review. Speech Recognition is the process of recognizing what is spoken in the context (questioned sample) by the speaker rather than who is spoken. But in the case of Speaker Identification, the expert try to find out who has uttered the given text by comparing the voice in the questioned sample with the given sample voice of the suspect(s). Understanding the characteristics of embedded noise structure associated with the recorded speech samples received for Forensic Speaker Identification is one of the challenging tasks to be tackled up by the experts in the field. Considering human/machine interfaces as a major area of applications, it is obvious that the signal becomes more challenging as the acoustic environment becomes more complex and hostile. Unnecessary disturbance within the frequency band of the speech in a communication channel or in other words any undesired sound is generally termed as Noise .Based on the means by which it is produced, noise has been generally partitioned into Additive and Convolutional. Additive noise has been considered as those produced from energy-radiating source and can be further classified into time-correlated and time uncorrelated.
Pagination: 269p.
Appears in Departments:Department of Forensic Science

Files in This Item:
File Description SizeFormat 
01_title.pdfAttached File84.94 kBAdobe PDFView/Open
02_dedication.pdf100.84 kBAdobe PDFView/Open
03_certificate.pdf106.13 kBAdobe PDFView/Open
04_declaration.pdf102.64 kBAdobe PDFView/Open
05_acknowledgements.pdf113.24 kBAdobe PDFView/Open
06_preface.pdf97.22 kBAdobe PDFView/Open
07_contents.pdf127.74 kBAdobe PDFView/Open
08_list of tables.pdf95.5 kBAdobe PDFView/Open
09_list of figures.pdf106.1 kBAdobe PDFView/Open
10_chapter 1.pdf425.95 kBAdobe PDFView/Open
11_chapter 2.pdf181.52 kBAdobe PDFView/Open
12_chapter 3.pdf432.65 kBAdobe PDFView/Open
13_chapter 4.pdf1.03 MBAdobe PDFView/Open
14_chapter 5.pdf125.05 kBAdobe PDFView/Open
15_chapter 6.pdf113.69 kBAdobe PDFView/Open
16_references.pdf169.36 kBAdobe PDFView/Open
17_abstract.pdf55.75 kBAdobe PDFView/Open

Items in Shodhganga are protected by copyright, with all rights reserved, unless otherwise indicated.

Altmetric Badge: