Speech intelligibility in DECT telephones

Examensarbete för masterexamen

Please use this identifier to cite or link to this item: https://hdl.handle.net/20.500.12380/3302
Download file(s):
There are no files associated with this item.
Type: Examensarbete för masterexamen
Master Thesis
Title: Speech intelligibility in DECT telephones
Authors: Emanuelsson, Kristian
Norgren, Mikael
Abstract: The mobility that is made possible with a cordless handset introduces the issues of the handset being used in different environments concerning background noise. This Master thesis investigates what essential frequency regions that are most important for intelligibility within the telephone bandwidth when the handset is used as a receiver in industry noise. Listening tests were performed for the investigations. A measure called Speech Reception Threshold in Noise (SRTN) was chosen for the investigation of speech perception in the handset. These measures were concluded with Parameter Estimation by Sequential Testing (PEST). A Swedish sentence material was used for the measures and the background noise was binaurally recorded with a manikin head. Performing the listening test included a graphical computer interface developed in Matlab that let the test subject control the pace of the test procedure. The data analysis was performed in SPSS. The overall results show the attenuation of low to mid frequency content results in lower SRTN than attenuation of the high frequency content within the telephone bandwidth. The main causes for these results depend on two coinciding factors. The background noise in this study is stronger for lower frequency content and thus performs more masking of low frequency to mid frequency content in the speech signals. Furthermore,the frequency region up to 2000 Hz is important for the formants F1 and F2 in vowel sounds as well as in consonant sounds. The best results in SRTN were obtained for two speech programs with 6dB attenuation around the 1/3 octave bands ranging from 315-1000 Hz and 630-2000 Hz. The overall SRTN results in these speech programs were approximately 3 dB less compared to no attenuation,which resulted in -19 dB (SRTN). 6 dB attenuation around the 1/3-octave band centred at 2 kHz resulted in the lowest SRTN of about 6 dB less than with no attenuation. Low SRTN was also produced by 6 dB attenuation around the 1/3-octave bands ranging between 1600-3200 Hz, resulting in values of about 5 dB less than with no attenuation.
Keywords: Samhällsbyggnadsteknik;Civil Engineering
Issue Date: 2004
Publisher: Chalmers tekniska högskola / Institutionen för teknisk akustik
Chalmers University of Technology / Department of Applied Acoustics
Series/Report no.: Rapport E - Chalmers tekniska högskola, Avdelningen för teknisk akustik. : E04-14
URI: https://hdl.handle.net/20.500.12380/3302
Collection:Examensarbeten för masterexamen // Master Theses

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.