Search academic texts

Information × Registration Number 0220U102655, 0118U004607 , R & D reports Title Development of a technique for extracting speech signal from an acoustic stream by using a probabilistic-spectral method popup.stage_title Head Chorniy Sergii V., Registration Date 23-04-2020 Organization Hon. Prof. M.S. Bokarius Kharkiv Research Institute of Forensic Examinations popup.description2  The object of study is the acoustic flow. The purpose of the work is to develop a technique for isolating the speech signal from the acoustic stream by the probabilistic-spectral method and a software product for the automation of broadcasting in files created both by modern digital recording devices and by software products. Theoretical and practical results - the technique of speech signal extraction from the acoustic stream is developed by the method of complex use of probabilistic-spectral transformations and artificial intelligence (AI) methods, which allows to significantly reduce the volume of audio stream information that should be heard by a forensic expert. A software product has been created that can be used in expert audio recording studies to significantly reduce listening time. Novelty - the first method of speech signal separation from acoustic flow was developed by the method of complex use of the group of probabilistic-spectral transformations and methods of AI (machine learning). Product Description popup.authors Brendel Olga I Lutsenko Christina V Niculin Constantine V Roman Anna I popup.nrat_date 2020-07-03 Close