Cough sound recognition based on Hilbert marginal spectrum

Author

Shasha Le ; Weiping Hu

Author_Institution

Coll. of Electron. Eng., Guangxi Normal Univ., Guilin, China

Volume

03

fYear

2013

fDate

16-18 Dec. 2013

Firstpage

1346

Lastpage

1350

Abstract

Marginal spectrums of cough sound and non-cough sound are obtained through Hilbert-Huang Transform. Speech signals of cough sound and five different non-cough sounds, namely throat clearing, sigh voice, shouting voice, speech voice and laughter, are analyzed contrastively focusing on the characteristics of marginal spectrum. Then SECC is extracted for cough sound recognition. The paper mainly recognizes the characteristic parameters SECC and MFCC coefficient by using the Continuous Hidden Markov Model(CHMM). The recognition result shows that SECC parameter based on Hilbert marginal spectrum is more effective to distinguish the cough sound and non-cough sound.

Keywords

Hilbert transforms; hidden Markov models; speech processing; CHMM; Continuous Hidden Markov Model; Hilbert marginal spectrum; Hilbert-Huang transform; cough sound recognition; marginal spectrum; non cough sounds; shouting voice; sigh voice; speech signals; speech voice; throat clearing; Cepstrum; Character recognition; Hidden Markov models; Mel frequency cepstral coefficient; Speech; Speech recognition; Training; Hilbert Huang Transform(HHT); Mel Frequency Cepstrum Coefficient(MFCC); Sub-band Energy Cepstrum Coefficient(SECC); cough sound;

fLanguage

English

Publisher

ieee

Conference_Titel

Image and Signal Processing (CISP), 2013 6th International Congress on

Conference_Location

Hangzhou

Print_ISBN

978-1-4799-2763-0

Type

conf

DOI

10.1109/CISP.2013.6743882

Filename

6743882