| Autor: |
Bhar, Anirban, Bhar, Poulami, Bhattacharyya, Soumya, Afreen, Sana, Saha, Sourav |
| Předmět: |
|
| Zdroj: |
AIP Conference Proceedings; 2023, Vol. 2876 Issue 1, p1-7, 7p |
| Abstrakt: |
In this paper, we propose a system to detect violence in conversation based on audio analysis of speech with an aim of sentiment rating. In general, a speech-to-text based approach is widely used for the identification of sentiment in audio. But there is a lack of focus on the scope of acoustic analysis of conversation without considering the transcription to identify violent audio conversation. Vocal features can be explored effectively to find emotion specific patterns in the voice for understanding the human mood during conversation. The modern techniques of deep learning are showing promising performances in various automated identification tasks. The power of deep learning models drives us to deploy a deep learning-based framework on conversational audio speech to detect whether it is violent or not. The key stages of our work are speaker diarisation, voice segmentation, voice segment clustering, and modeling of acoustic vocal features to detect violent audio content. The experimental results demonstrate the effectiveness of the proposed model in identifying the presence of violent content in a given audio stream. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
