Question

Q) Explain the time and frequency domain characteristic properties of the voiced and unvoiced parts of the human speech signal

Answer 1

First we will how the Speech signals are generated.

Speech signal generation:

Speech is generated as one exhales air from the lungs while the articulators move. Sound production process can be recognized as a filtering process in which the vocal tract filter
is excited by a speech sound filter. The source either is noisy aperiodic, causing voiced speech, or is periodic, causing unvoiced speech.

The source of the periodicity for the former is found in the larynx where vibrating vocal cords interrupt the airflow from the lungs, producing pulses of air.

Generally the vocal tract generates some sounds directly. It is the source for obstruent like stop and fricative sounds.

Different phones primarily can be distinguished by their periodicity - voiced or unvoiced, spectral shape - mostly which frequencies have major power and duration - longer phones
are perceived as having greater stress.

Fig. Generation of speech for voiced and unvoiced sound

Noise Vocal Tract Filter Pulse train Speech FO Excitation

In voiced stops, vocal folds may continue to vibrate throughout the stop or start to vibrate right after the burst.

-Only difference between voiced and unvoiced stops is that the vocal folds are more widely separated during the vocal tract closure for unvoiced stops and start to adduct only at the release, hence the longer VOT for unvoiced stops.

Properties and Characteristics of Speech Signal:-
Speech signal have following inherent property :
1. In a linear manner it can be noticed that speech is a sequence of continually changing sounds.
2. The properties of the speech signal are highly dependent on the sounds that are produced.
3. The properties of the speech signal are highly dependent on the context in which the sounds are produced. It implies that way in which the sounds which generally
take place before and after the current sound. This consequence is called speech sound co-articulation and it is the result of the vocal mechanism anticipating following sounds while producing the current sound, thereby changing the sound properties of the current sound. Some of the parameter of vocal cords like the positions, shapes and sizes of the various articulators like teeth , lips, tongue, jaw,etc.

Time and Frequency Domain Characteristics of Speech:-
Analyzing speech in the time domain often requires simple calculation and interpretation.

Among the relevant features found readily in temporal analysis are waveform statistics, power and F0. The frequency domain, provides the mechanisms to obtain the most useful parameters in speech analysis.

Most models of speech production assume a noisy or periodic waveform exciting a vocal-tract filter. The excitation and filter can be described in either the time or frequency domain, but they are often more consistently and easily handled spectrally.