“School of Cognitive”

Back to Papers Home
Back to Papers of School of Cognitive

Paper   IPM / Cognitive / 13438
School of Cognitive Sciences
  Title:   Predict the gender of speakers using cepstrum analysis and supporte vector machine
  Author(s): 
1.  S. Tabrik
2.  M.R. Daliri
3.  M. Behroozi
  Status:   In Proceedings
  Proceeding: 2nd Neuroscience conference, December 18-20, 2013
  Year:  2013
  Supported by:  IPM
  Abstract:
1. Introduction The question of how the speaker voice represents the information about the speaker gender has been discussed in many scientific communities. Recently need to gender identification have increased in biometric security application, mobile and automated telephonic communication and the resulting limitation in transmission bandwidth, practical application of gender identification in many of scientific and industrial fields. 2. Materials methods In this study, speech samples (Apple-Car-Happy) were collected separately for male speakers�?? group and female speakers�?? group (10 female and 10 male). Within each group, each member spoke the same language and same word for recognition purpose. The data were recorded at the noisy environment. 3. Results In this study, we tried to predict the speaker�??s gender using various voice signal processing techniques and algorithms. Firstly, the recorded speech signals clustered into voiced/unvoiced signals by using a fuzzy c-means algorithm implemented in MATLAB (version 2010). Then, the pitch of the participant�??s voice is extracted using the Cepstrum analysis. The extracted pitches as features separated to two sets, train set and test set, then train set was applied for train the SVM classifier and our classifier was evaluated with test set. The classifier results were 78.44. Conclusion In this work, an automatic approach has been implemented to gender classification using speech processing of a speech signal by clustering it into unvoiced/voiced portion. Extracted pitches of a voiced portion via Cepstrum analysis were used for gender classification with SVM.

Download TeX format
back to top
scroll left or right