Determination of glottal closure instants from clean and telephone quality speech signals using single frequency filtering. (November 2020)

• Record Type: Journal Article
• Authors: Kadiri, Sudarsana Reddy; Yegnanarayana, B.
• Is Part Of: Computer speech & language. Volume 64(2020)
• Publication Date: 2020-11
• Subjects: Speech analysis -- Glottal activity -- Glottal closure instant (GCI) -- Spectral variance -- Single frequency filtering (SFF) -- Zero frequency filtering (ZFF); 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.2020.101097
• Languages: English
• ISSNs: 0885-2308

Abstract

Highlights: The proposed method uses filtering, thus avoiding the window effects of block processing and also method does not assume any knowledge of the average pitch period to derive the glottal activity features, such as the impulse-like excitation corresponds to the glottal closure instants (GCIs). The property of the impulse, i.e., impulse in the time domain results in flat spectrum in the frequency domain is exploited using spectral variance. We demonstrate the benefit of exploiting the impulse-like discontinuity events for GCI detection in clean and telephone quality speech scenarios, and the proposed method is compared with the several established methods reported in the literature. The results of the proposed GCI detection method is found to be significantly better than several established methods for telephone quality speech and comparable in the case of clean speech. Abstract: A new approach for determining the glottal activity from speech signals is presented in this paper. The approach is based on the use of single frequency filtering (SFF), proposed recently for voice activity detection. The variance (across frequency) of the spectral envelopes at each sampling instant is derived using the SFF of speech signal. The variance plot shows discontinuities corresponding to the impulse-like excitation within each glottal cycle. The proposed method uses filtering, thus avoiding the window effects of block processing, but the temporal resolution of the frequencies is still determined by the filtering process. The method also does not assume any knowledge of the average pitch period to derive the glottal activity features, such as the locations of the glottal closure instants (GCIs). The proposed GCI detection method is evaluated on five databases and compared with four existing methods on clean speech and telephone quality speech. From the experiments, it is observed that the performance of the proposed method is comparable to the existing methods on clean speech and significantly better on telephone quality speech.