Effect of SVD Based Processing on the Perception of Voiced and Unvoiced Consonants
Randhir Singh, Parveen Kumar Lehana · 2012
Abstract— Speech is a biomedical signal used by the human beings to communicate. It is generated by exciting the vocal tract from the impulses of the air coming from the lungs through the vocal cords. Sometimes, the speech generated may not be adequate for understanding or transmission. In that case, it is modified using the concepts of speech processing. In this paper the singular value decomposition (SVD) technique is used to process and the output are evaluated using informal listening tests for investigating its effect on perception. This technique may have applications in speech compression, speech enhancement, speech recognition, and speech synthesis. The speech signal in the form of vowels-consonant-vowel (VCV) was recorded for the six speakers (3 males and 3 females). These VCVs were analyzed using SVD based technique and the effect of the reduction in singular values was investigated on the perception of the resynthesized VCVs using reduced singular values. Investigations have shown that the number of singular values can be drastically reduced without significantly affecting the perception of the VCVs. The vocal cords are vibrating valve that caused due to the airflow from the lungs generating the audible pulses that form the laryngeal sound source which adjust the length and tension of the vocal cords for the fine pitch and tone. The articulators articulate and filter the sound from the larynx and also strengthen or weaken the larynx as a sound source. The combination of vocal cords with the articulators is capable of generating highly intricate arrays of sound. Normally the speech frequency ranges from 60-7000 Hz but in males it is 60-180 Hz and 160-180 Hz in females (3).The human speech production is shown in Fig.1.