<p>with the rapid improvement and perfection of Internet science and technology, the use of computer-assisted teaching systems in Chinese universities has become more and more common. This kind of sensor network mainly collects and analyzes all kinds of information that can be felt within the network, and sends the processing situation to the inspectors. Language recognition technology is also improving with the improvement of network technology and computer technology. In people’s real work and life, it is widely used. After installing the language recognition system in the smart home system, it will make people feel more comfortable and convenient in their daily lives. This paper gives experimental research on the fusion algorithm of multiple basic sensor data. Mainly algorithms such as recursive estimation method, adaptive weighting method, arithmetic average method and clustered data method, and the related algorithms have been improved and perfected. Through the simulation experiment, the effect of the improved algorithm under the conditions of accuracy and so on is tested. It is proved that the improved algorithm can reduce the frequency of redundant data transmission and reduce the consumption during data transmission. At the same time, it analyzes and processes the recognition and distinction of language features. Research and analyze the relevant data obtained in the experiment. Based on the difference between the obtained data, the teaching system in colleges and universities will be further perfected and improved.</p>