Research on Speech Separation Technology Based on Deep Learning Approach

Intelligent terminal is an emerging term, it is the product of the continuous development of Internet+ IT technology, in just a few years, various terminal applications supported by the mobile Internet have emerged one after another, subtly affecting people's daily life, has become an important part of people's happy life. With the convenience of intelligent applications and the characteristics of big data traffic, network data traffic has increased exponentially. Voice is one of the important carriers of language interaction, it is like a bridge to connec+t one side of the object and the other side of the object terminal, so more and more scholars begin to study intelligent speech, and the intelligent processing of speech has gradually become a hot spot in the market. As scholars introduce deep learning technology into the field of speech intelligent processing, it is possible for intelligent terminal services to process a large number of voices at the same time and can extract various effective information from input speech, such as emotions, meanings, keyword frequencies, and so on. However, in the process of practical application, speech is very susceptible to interference from various uncertain factors, there is "noise" in the process of voice propagation, which not only increases the difficulty of the work of the speech intelligent processing algorithm, but greatly reduces its performance, but also affects the accuracy of the processing results to a certain extent, which makes the entire speech intelligent processing process have great space for development and exploration, so the focus of scholars' research begins to gradually bias towards how to perform speech separation. Eliminate noise interference and achieve ideal application results in real-world scenarios. The research and development of speech separation technology is related to important applications in many fields, and speech separation plays the dual role of auxiliary tools and effective tools.