Generally, CASA also includes non-sinusoidal modeling based analysis system. Many systems do not even aim to acheive perfect separation when it comes to just understanding the auditory scene. It can deal with every day sound recorded on the street, in a resaurant etc. with the aim to analyse the recorded signal into auditory objects and understand what is going on in the scene. Prediction driven CASA proposed by Dan Ellis used transient clicks(for brief bursts of energy) and wefts(for wideband periodic enrgy) to represent the signal (Ellis 1993) [8]. They were then grouped according to prediction from competing hypotheses. The system not only used the bottom-up approaches, refering to the process of identifying each partial components and then group them according to the psychoacoustic cues, but also predicted to-down from the acoustic cue grouping principles to determine each component.