Speech Disability Threshold Determination by Graphical and DSP Techniques

Abstract

In this paper, an attempt is made to graphically distinguish the normal and pathological cases of speech disabled children. The digitized acoustic signal from children of age range 5 to 8 years are captured and subjected to Linear Predictive Coding (LPC), FFT and CEPSTRUM processes. The output data obtained from LPC, FFT and CEPSTRUM are analyzed for four cases viz, one normal and three pathological cases of varying degrees of severity. The usefulness of the proposed technique is further illustrated as the locations of poles are found to have shifted within the unit circle. The results are summarized in the form of three indices namely co-relation index of FFT signal, co-relation index of CEPSTRUM and the order of prediction in LPC for severity threshold determination. The results from FFT and CEPSTRUM are found to match with that of the LPC technique favorably.

Keywords

• CEPSTRUM
• Correlation
• FFT
• Linear predictive coding and speech disability