An Efficient Speaker Diarization using Privacy Preserving Audio Features Based of Speech/Non Speech Detection

Apply

An Efficient Speaker Diarization using Privacy Preserving Audio Features Based of Speech/Non Speech Detection

Journal Title: INTERNATIONAL JOURNAL OF COMPUTER TRENDS & TECHNOLOGY - Year 2014, Vol 9, Issue 4

Abstract

Privacy-sensitive audio features for speaker diarization in multiparty conversations: i.e., a set of audio features having low linguistic information for speaker diarization in a single and multiple distant microphone scenarios is a challenging research field in now-a-days. Existing system used a supervised framework using deep neural architecture for deriving privacy-sensitive audio features. In proposed system Patterns of speech/nonspeech detection (SND) is utilized for privacy-sensitive audio feature to capture real-world audio. SND and diarization can then be used to analyze social interactions. In this research privacy preserving audio features has been investigated instead for recording and storage that can respect privacy by minimizing the amount of linguistic information, whereas achieving modern performance in conversational speech processing tasks. Certainly, the main contribution of the proposed system is the achievement of state-of-the-art performances in speech/nonspeech detection and speaker diarization tasks using such features, which we refer to, as privacy-sensitive. In addition a comprehensive analysis of these features has been provided for the two tasks in a variety of conditions, such as indoor (predominantly) and outdoor audio. To objectively evaluate the notion of privacy, the proposed system use automatic speech recognition tests, with higher accuracy in either being interpreted as yielding lower privacy.

Authors and Affiliations

S. Sathyapriya , A. Indhumathi

Keywords

EP ID EP94203
DOI -
Views 78
Downloads 0