• developed a standard Baum-Welch training module

• integrated it into ISIP public domain speech recognition system

• features included are multiple Gaussian mixture components training, the use of phone and word level transcription and the capability of handling cross-word triphones