Principal component analysis of the spectrogram of the speech signal: Interpretation and application to dysarthric speech. (January 2020)

• Record Type: Journal Article
• Authors: Kacha, Abdellah; Grenez, Francis; Orozco-Arroyave, Juan Rafael; Schoentgen, Jean
• Is Part Of: Computer speech & language. Volume 59(2020)
• Page Start: 114
• Page End: 122
• Publication Date: 2020-01
• Subjects: Principal component analysis -- Spectrogram -- Long-term average spectrum -- Dysarthric speech; Speech processing systems -- Periodicals; Automatic speech recognition -- Periodicals; Computers -- Periodicals; Linguistics -- Periodicals; Speech-Language Pathology -- Periodicals; Traitement automatique de la parole -- Périodiques; Reconnaissance automatique de la parole -- Périodiques; Automatic speech recognition; Speech processing systems; Electronic journals; Periodicals; 006.454
• DOI: 10.1016/j.csl.2019.07.001
• Languages: English
• ISSNs: 0885-2308

Abstract

Highlights: Interpretation of the principal components of the spectrogram of the speech signal. The first principal component of the spectrogram of continuous speech is highly correlated with the long-term average spectrum. The second principal component is the difference of two weighted sums of frame spectra reporting open and close vowel frame spectra respectively. Principal component analysis of the spectrograms of speaker monologues is used to document articulatory deficits in Parkinson speakers. Statistically significant differences between Parkinson and control speakers are observed. Abstract: The article concerns the interpretation of the principal components of the spectrogram of the speech signal and its application to the description of dysarthric speech. Each principal component is a linear combination of the frame spectra. We show that the first principal component of the spectrogram is closely related to the long-term average spectrum (LTAS) and the second principal component is the difference of two weighted sums of frame spectra reporting open and close vowel frame spectra respectively. We investigate articulation deficits in dysarthric speakers via cues obtained from principal components of the spectrogram of connected speech because long-term average spectra have been claimed to inform about speaker settings of the vocal tract.