Publication detail

Utilizing MFCC for Voice Intensity Determination

YOUNES, D. ŠTEFFAN, P.

Original Title

Utilizing MFCC for Voice Intensity Determination

English Title

Utilizing MFCC for Voice Intensity Determination

Type

journal article - other

Language

en

Original Abstract

This paper aims to classify speech signals according to their intensities into 3 groups (silent speech, normal speech and very loud speech). In this work, we have used a number of speech recordings, which were stored in a WAV format. These recordings are divided into groups with three different intensities, with each of them at three different distances from the point of entry (microphone). Mel Frequency Cepstral Coefficients and k-Nearest Neighbor classifier based algorithm was tested and implemented in Matlab environment. The intensity classification accuracy achieved was about 85%.

English abstract

This paper aims to classify speech signals according to their intensities into 3 groups (silent speech, normal speech and very loud speech). In this work, we have used a number of speech recordings, which were stored in a WAV format. These recordings are divided into groups with three different intensities, with each of them at three different distances from the point of entry (microphone). Mel Frequency Cepstral Coefficients and k-Nearest Neighbor classifier based algorithm was tested and implemented in Matlab environment. The intensity classification accuracy achieved was about 85%.

Keywords

voice intensity, MFCC, cepstrum, k-nn algorithm

RIV year

2010

Released

22.09.2010

Location

Sofia, Bulgaria

Pages from

162

Pages to

164

Pages count

3

BibTex


@article{BUT50165,
  author="Dina {Younes} and Pavel {Šteffan}",
  title="Utilizing MFCC for Voice Intensity Determination",
  annote="This paper aims to classify speech signals according to their intensities into 3 groups (silent speech, normal speech and very loud speech). In this work, we have used a number of speech recordings, which were stored in a WAV format. These recordings are divided into groups with three different intensities, with each of them at three different distances from the point of entry (microphone). Mel Frequency Cepstral Coefficients and k-Nearest Neighbor classifier based algorithm was tested and implemented in Matlab environment. The intensity classification accuracy achieved was about 85%.",
  chapter="50165",
  journal="Electronics",
  number="2",
  volume="4",
  year="2010",
  month="september",
  pages="162--164",
  type="journal article - other"
}