Research And Implementation Of Speech Signal Processing Method On Government Information Platform

With the attention to “intelligent government” and the rapid development of artificial intelligence,people put forward to higher requirements for the timing,degree and effectiveness of the government information platform.The operation and maintenance of the government information platform require a large amount of text inputs,and manual pinyin input method seriously affects the input speed of Chinese characters,thus affecting the related work process.As a technology that can replace the traditional pinyin input,speech recognition technology has great potential in the government information platform due to its accuracy and timeliness.At present,it is not mature to apply speech recognition technology to the application system of government information platform.Therefore,this paper researches the realization of the speech signal processing method in the government information platform.According to the functional requirements of speech signal processing of government information platform,this paper focuses on the speech signal processing method and implementation of government information platform,especially research and analysis are carried out on speech enhancement algorithm,endpoint detection algorithm and speech recognition algorithm.According to the author’s actual work,a targeted speech signal processing method is proposed,and the implementation and testing of the voice signal processing software of the government information platform are completed.First of all,the spectral subtraction method is used to enhance the input noisy speech signal,so as to improve the signal to noise ratio.Then,aiming at the problem of signal loss in the traditional double-threshold endpoint detection algorithm,an improved double-threshold endpoint detection algorithm is proposed that can effectively solve the problem of signal loss detection.Then,the MFCC coefficients of the speech signal based on the auditory perception frequency are used to extract the speech signal features.Finally,aiming at the local optimal solution in the BP neural network speech recognition algorithm,the BP neural network based on the particle swarm optimization is studied,and the weights and thresholds of connections between layers within the network are optimized through the particle swarm,which can reduce the running time and improve the recognition accuracy.Meanwhile,the software implementation method of the government information platform is further studied,and the proposed algorithm is implemented by using Python environment and the related functional modules are tested.The test results demonstrate that the proposed method can implement the speech recognition function of the government platform.