Timbre is one of the most important features in qualifying the sound of a vocalist. This study is focused on identifying the timbre characteristics of vocal sound signals of eminent vocalists in Hindustani music and classifying their voices accordingly. In our study we have considered the following timbral parameters, viz. Irregularity of the partials, Brightness, Tristimulus1 (T1), Tristimulus2 (T2), Spectral Centroid, Jitter or Irregularity of the frequencies of the partials, Shimmer or Irregularity on the amplitude evolution of the partials, Odd parameter and Even parameter. In this study we have taken the sung voices of three male and three female singers. Each of them has sung four ragas. Due to bad rendering one signal was rejected. Therefore, twenty-three (23) signals of equal durations were considered for analysis. Since the objective of the work is to classify the voices with similar timbre, raga wise classification was not attempted. The study was concentrated on the singer wise data for each male and female singer. For each singer these parameters were measured, studied and compared separately among the three male singers named as A, B and C and three female singers named as D, E and F and classification of these singers were attempted. Findings of these parameters were compared statistically based on Minkowski distance to determine nearest neighbor or nearest points. A and B among male singers and E and F among female singers were found to be nearest neighbor of each other.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Timbre Based Classification of Hindustani Vocalists

  • Aditi Mandal,
  • Anirban Patranabis,
  • Kaushik Banerjee,
  • Argha Deb,
  • Ranjan Sengupta,
  • Dipak Ghosh

摘要

Timbre is one of the most important features in qualifying the sound of a vocalist. This study is focused on identifying the timbre characteristics of vocal sound signals of eminent vocalists in Hindustani music and classifying their voices accordingly. In our study we have considered the following timbral parameters, viz. Irregularity of the partials, Brightness, Tristimulus1 (T1), Tristimulus2 (T2), Spectral Centroid, Jitter or Irregularity of the frequencies of the partials, Shimmer or Irregularity on the amplitude evolution of the partials, Odd parameter and Even parameter. In this study we have taken the sung voices of three male and three female singers. Each of them has sung four ragas. Due to bad rendering one signal was rejected. Therefore, twenty-three (23) signals of equal durations were considered for analysis. Since the objective of the work is to classify the voices with similar timbre, raga wise classification was not attempted. The study was concentrated on the singer wise data for each male and female singer. For each singer these parameters were measured, studied and compared separately among the three male singers named as A, B and C and three female singers named as D, E and F and classification of these singers were attempted. Findings of these parameters were compared statistically based on Minkowski distance to determine nearest neighbor or nearest points. A and B among male singers and E and F among female singers were found to be nearest neighbor of each other.