The effect of the use of white noise for masking stuttered speech reconstruction

Abstract

Stuttering can be defined as the unintentional disruption in the normal flow of speech by dysfluencies, which include repetitive pronunciation, prolonged pronunciation, blocked or stalled pronunciation at the phoneme or the syllable level. The effect of noise  masking on the reconstructed stuttered speech is the focus of this study. This study aimed at finding out the effects of white noise  masking on the reconstruction of stuttered speech. Three stuttered words; anniversary, department and sales were masked with 5dB  white noise. LPC analysis – synthesis was used for the speech reconstruction, while Welch power spectral density (PSD) estimates was  used in evaluating the speech signals in frequency domain. The algorithm effectively recreated the speech samples via reconstruction.  The dominant peaks from about 2 kHz were modulated by the masking noise. As such, all the repetition in the noise masked region have  reduced power, while the lowest frequency points also had its power increase for the three stuttered words considered. The added white  noise as a masking noise thus effectively reduced the repetitions and by extension the stuttering in the speech.