The Effect of Filtered Speech on Speech Identification Scores of Young Normal Hearing Adults

Abstract

Filtered speech material would help to understand the importance of different spectral energy in perception of speech. Spectral energy of different speech sounds varies due to variation of the fundamental frequency (F0), first formant (F1) and second formant (F2) values. These are essential cues to understand speech. The F1 and F2 values are determined by the tongue height and tongue advancement respectively. These formants can alter due to the change in volume or size of the oral structures and also the language used. Hence, the present study aimed at assessing the speech identification scores for four low-pass (800, 1200, 1500 & 1700 Hz) and four high-pass (1700, 2100, 2500 & 3000 Hz) cut off frequencies filtered speech in 30 young normal hearing adults in the age range of 17- 25 years. The spectral modifications were done using Adobe Audition Software for the phonemically balanced words in Kannada developed by Yathiraj and Vijayalakshmi (2005). The speech identification scores were determined at all the eight cut off frequencies. Results revealed that there is lowering of the cut off frequencies at which 70% speech identification scores are obtained for low-pass filtered stimuli in Kannada compared to English (Bornstein, Wilson & Cambron, 1994). This difference could be because of the low F1 values and higher low frequency spectral energy in Kannada spoken language compared to English language. There is no difference in high-pass cut off frequency filtered speech. The results suggested that the spectral information between 1200 Hz and 2100 Hz is more important in perception of speech in Kannada language and also highlight the selection of appropriate cut off frequency for filtered speech for clinical use.