Evaluation of English Speech Interaction Quality Based on Deep Learning

In order to improve the traditional computer English pronunciation quality assessment method, this paper uses the spoken English of university students in China as the research object, and examines various measures such as intonation accuracy, speaking rate, rhythm, and intonation. Intonation is measured by Mel-frequency cepstral characteristic expression, speech rate according to speech time, rhythm is evaluated by short-term power and pair variation index, and intonation is evaluated as the base frequency. Based on the evaluation indicators of English pronunciation quality, this article develops an English pronunciation evaluation model based on DBN. Experiments were conducted on the evaluation performance of the model, and the test results showed that the recognition rate of the DBN model developed in this paper was 96.65%, which is better than other models. In addition, consistency experiments were conducted between machine evaluation and manual evaluation, the total value of machine and manual evaluation is 87.5%, the environmental similarity level is 100%, and the Pearson correlation coefficient is 0.722, indicating similarity between machine evaluation and manual evaluation.