Acoustic Segment Modeling using Spectral Clustering techniques

Acoustic models plays a very important role in many speech applications like speech recognition, spoken term detection, topic classi cation, language identi cation etc. Acoustic segment modeling (ASM) is a method to build acoustic models based on acoustic similarities of speech segments.\ud Typically, the standard process of training acoustic models is supervised which has attained great success in past. The issue with supervised techniques is the availability of large amount of transcribed data and language-speci c knowledge. Transcribing the speech data is time consuming,tedious and demands a lot of expert human labour, and hence expensive