Forensic intelligibility enhancement of recorded speech in high noise environments

2016 Fiscal Year Final Research Report

• Project/Area Number: 24710195
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Social systems engineering/Safety system
• Research Institution: National Research Institute of Police Science
• Principal Investigator: Makinae Hisanori 科学警察研究所, 法科学第四部, 主任研究官 (20415441)
• Research Collaborator: SATO Masamichi
• Project Period (FY): 2012-04-01 – 2017-03-31
• Keywords: 信頼性工学 / 裁判科学

Outline of Final Research Achievements

In a forensic situation, since speech samples are usually recorded in highly noisy environments, for example, in a store or a car where background music is played, the signal-to-noise ratio (SNR) of these samples is severely degraded. We propose speech enhancement methods effective to these samples in this study. In some proposed methods, source signals of the noise were used as reference signals, because real time processing is unnecessary for forensic purpose and the source signals used for background music is easily available. And non-negative matrix factorization (NMF) or sinusoidal modeling was used for signal representation. Experiment showed the effectiveness of the proposed methods to the severely degraded signals. In addition to these methods based on signal processing technique, preliminary examination was carried out to investigate the availability of speech recognition technology.

Free Research Field

法科学