Enhancement of Spanish Oesophageal Speech vowels using coherent subband modulator Kalman filtering

Article type: Research Article

Authors: Ishaq, Rizwan^* | García Zapirain, Begoña

Affiliations: Deustotech-LIFE, University of Deusto, Bilbao, Spain

Correspondence: [*] Corresponding author: Rizwan Ishaq, Deustotech-LIFE, University of Deusto, Bilbao, Spain, E-mail:rizwanishaq@deusto.es

Abstract: This paper proposes an Oesophageal Speech (OES) enhancement method, based on Kalman filtering. The Kalman filter is applied to modulators of OES frequency subbands instead of the fullband signal. The OES frequency subbands are decomposed into modulators and carriers components using coherent demodulation. In comparison with fullband Kalman filtering and pole stabilization, the proposed technique shows better results. The system performance is evaluated objectively and subjectively using the Harmonic to Noise Ratio (HNR) and Mean Opinion Score (MOS) respectively. Results have shown that Kalman filter in subband modulators processing is robust and efficient, improving the HNR by 4 to 5 dB for all Spanish vowels.

Keywords: Alaryngeal speech, Kalman filtering, filterbank, synthesis/analysis, modulation frequency domain

DOI: 10.3233/THC-161129

Journal: Technology and Health Care, vol. 24, no. 2, pp. 201-213, 2016