Research on the Evaluation of Chinese Articulation Using Machine Learning and Algorithm Optimization

This design uses SPCE061A, SPLC501 and keyboard module to realize speech recognition system. The system realizes the functions of adding, deleting and searching basic information. The data saved in SPR4096 was recognized by voice. A steady state interference noise for Chinese speech audibility detection based on pink noise is proposed, and its frequency is consistent with the speech features recorded in anechoic chamber. The corresponding training samples were obtained by convolution of the Chinese speech audibility test word list with the equivalent binaural chamber impulse response. It is then tuned to a specific signal strength. The objective evaluation of the intelligibility of Chinese speech is done by audiologists. It is found that factors such as reverberation time, signal-to-noise ratio and pronunciation speed play an important role in the intelligibility of Chinese speech. Compared with reverberation time and speech frequency, SNR plays a more significant role in the intelligibility of Chinese speech. Compared with conventional methods, this method has better recognition performance and speed.