Hybrid pitch extractor for speech utterances

S. KWONG, Q. H. HE, K. F. MAN

Research output: Journal PublicationsJournal Article (refereed)peer-review

Abstract

This paper presents a new pitch detection algorithm similar to the Semiautomatic Pitch Detector (SAPD) in the sense that it initially calculates the pitch period with a time-domain principle cycle extractor. However, it differs from SAPD in that the pitch periods are not only calculated on time-domain, but are assured with an error detector. If there is any ambiguity, the pitch period will be re-calculated with the CEP algorithm. A `pitch selection logic' is used to select the pitch period from the two estimates. Experimental results show that the algorithm exhibits good performance in the five measurements proposed in [2]. The new algorithm runs at half the speed of the Data Reduction Method (DARD), but faster than the Modified Autocorrelation Method (AUTOC) and the Cepstrum Method in the order of four and eight times, respectively.
Original languageEnglish
Pages (from-to)313-319
Number of pages7
JournalInternational Journal of Modelling and Simulation
Volume20
Issue number4
DOIs
Publication statusPublished - 2000
Externally publishedYes

Keywords

  • Pitch period
  • principle cycles
  • fundamental frequency
  • pattern recognition

Fingerprint

Dive into the research topics of 'Hybrid pitch extractor for speech utterances'. Together they form a unique fingerprint.

Cite this