Speech signal analysis as an alternative to spirometry in asthma diagnosis: investigating the linear and polynomial correlation coefficient

Abstract

Speech production involves the vibration of the vocal cords. Voice changes will occur in respiratory diseases such as asthma due to the inflamed lung airways, which is part of the vocal tract. Spirometry is a well-known technique employed in diagnosis of asthma to give information on patient pulmonary function. The purpose of this research was to investigate the correlation between Forced Expiratory Volume to Forced Vital Capacity (FEV1/FVC) ratio obtained from spirometry and Harmonics-to-Noise Ratio (HNR) obtained from human speech, in order to determine whether speech analysis could be an alternative to spirometry in diagnosing asthma. Spirometry data was obtained from 150 subjects, who were asthmatic patients attending the Korle-Bu Teaching Hospital, Ghana. Speech data consisting of the vowel sounds /a:/,/e:/, /ɛ:/, /i:/,/o:/, /ɔ:/,/u:/ and phrase “She sells”, was also recorded from the subjects. 33 samples were selected and analyzed to generate speech parameters with Praat software. Correlation was established between HNR from the speech signals and spirometry data FEV1/FVC. The highest correlation coefficient was observed between HNR and vowel sound /ɛ:/ (42.08%). In conclusion, among the other speech vowels and phonemes, HNR of /ɛ:/ sound showed the most promise to being a suitable marker in using speech as an alternative to spirometry in asthma diagnosis.