Research progress and trend analysis of speech recognition technology using CiteSpace and computer neural network

Speech recognition as a natural way of humancomputer interaction has been gradually applied to People's Daily life. In this paper, more than 40,000 literatures related to speech recognition from 2001 to 2020 in the core collection database of Web of Science were analyzed and processed visually by using the method of bibliometrics and visualization analysis software CiteSpace. We performed a visual network analysis of the network of partnerships and academic impact of speech recognition technology at the level of countries, research institutions and authors. We also analyze the research field, research hotspot and research frontier of speech recognition technology, and summarizes the research status and development trend of the global research. And in view of the three practical problems of noise and far-field interference existing in the current technology application, the time consuming of deep learning model training and the limited application scenarios of speech recognition, we propose three improved methods: using multi-channel microphone array technology for speech processing, training end-to-end recognition model and using optimized network structure.