The Emotional Variability Voice Classification And Identification Of Research

The classification and identification research of emotional variation speech as anemerging area in recent years. In the past, the speech system software to focus only onthe content of the speech expression is accurate, but not speech emotional factors intoit. And usually determine the emotional category by facial expressions, gestures, bloodpressure, heart beat and speech, etc, but speech has taken on a pivotal role. Therefore,many researchers focus shifted to the study of the speech of emotions.This paper mainly studies two problems: one is the classification of emotionalvariation speech, that is, to distinguish the speech belongs to which emotion; another isthe identification of emotional variation speech, also is, to identify the speech toidentify which sentence. The main work is as follows:(1)Designed and constructed contains natural, happy, sad and angry four types ofspeech emotions corpus, and for experimental research,which made theoreticalknowledge was more convincing.(2)Kaiser window was used. The function can be free to choose the main lobewidth and side lobe attenuation, effectively improve the recognition rate.(3) The characteristic parameters of emotional variation speech were analyzed indetail, and extracted emotional variation speech features, including energy, amplitude,zero crossing rate, pitch frequency, formant, linear prediction cepstrum coefficient,Mel-frequency cepstral coefficients and their derivative features.(4)Using average learning subspace method for the classification of emotionalvariation speech was proposed, the method was effective to improve the classificationaccuracy of the system, and the computing speed is faster.(5)Difference subspace method for the identification of emotional variationspeech was used, and the idear of first classified and then identify was preposed, which was classified for emotional variation speech after identify the contents of the speech.This will not only determine the speech emotion, but also to identify the contents of thespeech.