ABSTRACT
There are a number of ways a segment's length can be measured. For example, one could use the number of samples, but using these units would result in a measure that depended on the sampling frequency and the size of the frame. To avoid any confusion, this chapter measures all segment lengths in units of time. In the chapter, weak-voiced is lumped into the unvoiced class. Also, the voiced portions of the transitions are lumped into the voiced class and the unvoiced portions are lumped into the unvoiced class. Combinatory techniques form the basis on which many of the statistical analysis methods used throughout this chapter are founded. Since the chapter aims to describe a method used to classify usable speech segments using an estimate of the Target-to-Interferer Ratio (TIR), it would be prudent to have a statistical model of the segmental TIR signal. It develops a statistical model of the segmental TIR for cochannel speech.
