| Summary: | In this thesis, an android application aimed at studying the behaviour of an individual by analysing only the audio signal is developed. Normally, speech analysis or synthesis is done for an individual, but this will differ dramatically when it is dealt in a socio gathering because the number of voices is multiple. This application will help in self-analysis for an individual regarding his/her portrayal of emotion and social behaviour.
Moreover, in this study, voice prosody is used in communicating mannerism of an individual in the group. Communicating the behaviour is termed as “sociofeedback” in the entire report. The underlying concept in extracting the speech parameters is linked in the pitch (i.e. perceived frequency) and the actual frequency, Mel frequency cepstral coefficients along with the sound pressure level allows emulating the entire hearing system of human beings. The algorithm developed is sensitive to small changes in pitch variation at lower frequency ranges than to the higher frequencies. This steepness is exploited to track the inflections that occur in the emotions. These features along with few other derived features are used for classifying the input audio signal from an individual. Since the Google Glass was unable to handle the complex processing of audio data and the classification algorithm, only a couple of the features were possible to retrieve. Hence, the development of a full fledge application involving all the digital signal processing on the audio data along with a classification algorithm to process the extracted attributes of the signal was implemented on Android Phone.
|
|---|