Mixture Speech Separation Based On Computational Auditory Scene Analysis

The speech separation technology based on computational auditory scene analysis has a very wide range of applications in the field of artificial intelligence and machine perception and automatic speech separation,and it has gradually become one of the hot,especially in noisy environment, the speech signal separation is the most difficult.In this paper, based on the theory of computational auditory scene analysis,the mixed speech signal separation in noisy environment is studied.Mainly aimed at the existing problems in the original mixed speech separation system which uses of interaural time difference and interaural intensity difference as speech separation cues, further research and improvement are made.First of all, this paper presents a pitch period characteristic and interaural time differences and interaural intensity difference feature combination algorithm of separation,and the design of the dual masking model.The improved algorithm uses two types of speech separation cues, analyzes the mixed speech signals from two different angles, and then realizes the pure separation of the target speech by the double masking.Secondly, the original system can not completely mask sounds interference, this paper adds a separation method uses the pitch periodicity characteristics as speech separation cues,at the same time, a reasonable initial masking model is designed to remove the noise from the mixed speech,and combined with the follow-up of the secondary masking model, to achieve a more comprehensive cover, remove the interference of sound more thorough effect.Again, aiming at the problem that the original system can not accurately separate the speech of relatively large time delay,this article in the part based on interaural time difference and interaural intensity difference feature of mixed speech separation, the re-definition and improvement of secondary masking model are made, to enable the system can separate the target more clear and precise separation of any one voice signal.Finally, a lot of experiments are carried out to evaluate the performance of theimproved system, and compared with the original speech separation system,it can clearly reflect the effectiveness and superiority of the improved system.