ANEW TECHNIQUE FOR THE ESTIMATIONOFJITTERAND SHIMMER OF
Voiced Speech Signal · 2006
This paper presents a new technique fortheestimation of jitter andshimmer exploiting thedetected value ofthepitchperiod andamplitude ofvoiced speech signal degraded by white noise. Perturbations oftheglottis cycle duration (itter) andintonation (shimmer) are,respectively, theirregularities of pitch-period andenergy. We propose a modified circular autocorrelation function (MCACF)of thenormalized hamming-windowed pre-filtered speech. A weighted andpitchharmonic MCACF isutilized toeffectively extract pitch-period forjitter computation. Tocarryoutshimmer analysis, peak-topeakamplitude ofsignal around theestimated pitch-period is optimally obtained bymatching thewindowed pre-filtered speech framewith a shifted impulse train. Forperformance evaluation vowels, suchas la, e,i,o,ul, are usedfromthe TIMiT database. Theconventional algorithms manifest notonly thedistortion ofpitch contoursbutalsoprovides elevated values ofthejitter andshimmer that can beinterpreted as the presence ofanomalies ofthevocal tract, theglottis or the neuro-mechanism. Onthecontrary, forbothclean andnoisy speech, theproposed method results ina comparatively smooth pitch contour aswell assmall values ofthejitter andshimmer confirming the presenceofa normal voicewithgoodstability andperiodicity.