In order to further improve the human-machine interaction effect of current English service-oriented robots, the author proposes an English intelligent dialogue system based on deep learning algorithms using speech recognition as the basic method. By optimizing the feature parameter extraction method and speech recognition model of the interaction system, combined with corresponding module design, the performance of the interaction system has been improved to a certain extent. The simulation results show that compared with other feature parameter extraction algorithms, the LPMFCC feature parameter extraction algorithm in this study has a higher recognition rate, reaching 87.4%. Compared with the pre-improved HMM model, the improved HMM model proposed by the author can complete training faster and with lower training errors, requiring only 4 training sessions to complete the training. The English intelligent dialogue human-computer interaction system proposed by the author can achieve good performance and achieve good human-computer interaction effects, which has a certain reference value for practical design.