Feature analysis of scream sound and its application to scream detection system

2015 Fiscal Year Final Research Report

• Project/Area Number: 25750137
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Social systems engineering/Safety system
• Research Institution: Osaka Electro-Communication University
• Principal Investigator: Hayasaka Noboru 大阪電気通信大学, 情報通信工学部, 講師 (50554573)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: 悲鳴検出 / 帯域制限 / スペクトルエントロピー

Outline of Final Research Achievements

Recently, as crimes have become more diverse and increasingly sophisticated, our awareness of security systems has increased. The most typical kind of security system is a security camera, but there are various issues in that there are situations in which they cannot be used, such as from consideration of privacy or in darkness. For that reason, security systems that utilize sound information have been proposed. This report relates to scream detection system that focuses on screams within various sounds, and proposes techniques increasing its noise robustness. More specifically, we use spectral entropy, which expresses the difference in frequency distribution between a scream and noise, as a feature, and also construct a model from screams with superimposed noise. In scream identification and scream detection experiments under noisy environments, the proposed method could be improved the equal error rate under all noisy environment, in comparison with the conventional method.

Free Research Field

音声信号処理