A study on voice sobriety test algorithm in a time-frequency domain

Abstract

Voice is an important tool to identify individuals. In recent years, the voice can determine personality, age, and health status as well as physical conditions such as height and stature. The purpose of this paper is to develop an algorithm to determine whether when one is intoxicated by using the voice; characteristic parameters of the intoxicated voice signal were extracted by comparing the voices before and after drinking using several speech analysis methods. First, the change was observed in the body, glottis and vocal tract after drinking in order to examine the influence of drinking on phonation. In addition, differences of voice signals of a variety of the speakers were examined before and after drinking in order to find characteristic parameters for intoxicated voice signal that enable a judgment about whether or not one is drunk. The result of this study is expected to prevent drinking and driving, if applied to the development of the intoxication judgment system. © 2013 SERSC.