Affiliations: DeustoTech-Life, University of Deusto, Bilbao, Spain
Correspondence:
[*]
Corresponding author: Ibon Oleagordia-Ruiz, DeustoTech-Life, University of Deusto, Avda. de la Universidades 24, P.C.: 48007, Bilbao, Spain. Tel.: +34 94 413 90 00; Fax: +34 944 13 91 01; E-mail:ibruiz@deusto.es
Abstract:
BACKGROUND: This manuscript presents oesophageal speech enhancement.
Patients who have undergone a laryngectomy as a result of larynx cancer,
that is, laryngectomees, have communication problems. Due to removal of the
larynx, oesophageal speech has extremely low intelligibility. OBJECTIVE: Thus, it is necessary to process acoustical parameters, such as
Harmonic to Noise Ratio (HNR) in order to increase intelligibility. METHODS: The research focused on oesophageal Spanish /a/ phoneme
improvement. In order to enhance oesophageal speech two techniques were
applied: Kalman filtering and an algorithm which stabilizes the vocal tract
poles. Speech enhancement was measured using the MDVP tool. The oesophageal
voice database was compiled with the help of the local association of
laryngectomees. RESULTS: The results show an average improvement of 4.2 dB in the HNR.
Statistically, differences on average between the original and processed
voices, (p < 0.001) for HNR parameter were proven to be significant and we
therefore conclude that voice quality was improved due to evidence of a
higher HNR on average. CONCLUSIONS: As a conclusion, the study confirms oesophageal voice
enhancement since speech parameters are closer to the normal average range.
Subjectively, the oesophageal breathing noise is reduced substantially, as
is reflected in the MOS test.
Keywords: Medical informatics, rehabilitation engineering, signal processing, oesophageal speech, acoustical parameters enhancement
