Research And Implementation Of Distributed Intelligent Voice Switch Array

With the continuous development and advancement of technology,speech recognition technology has also been developed accordingly.The performance of speech recognition has also been further improved,the recognition rate is getting faster and faster,and the recognition rate is getting higher and higher.As a convenient,fast and efficient human-computer interaction method,the speech recognition system has attracted more and more attention.However,in practical application scenarios,speech recognition is often affected by the recognition environment and the training environment,resulting in a sharp decline in recognition performance.Therefore,improving the recognition rate and recognition speed of speech recognition in noisy environments has attracted more and more attention.This is also the key to the application of speech recognition systems out of the laboratory to people's lives.In order to make up for the shortcomings of the traditional methods,this paper combines distributed technology and wireless networking technology to develop a switch array design with priority,which can be widely used in home environments.The switch array uploads the voice signal to the server through the gateway to perform corresponding voice signal processing,which greatly improves the processing efficiency.At the same time,considering that there are many noises in the use of the scene,the effective speech signal is usually mixed with a large amount of environmental noise.This paper uses five microphones for noise reduction.When multiple microphone arrays exist at the same time,the delay estimation algorithm can be accurately determined to a specific microphone array,and the order of the voice signal sources to reach the microphones can be determined to locate the sound source position to ensure the consistency of the control target object,and Wi-Fi can be used to synchronize the other switches to determine whether the collected command is the first command or the misjudgment caused by the reflection,and the corresponding operation is performed.At the same time,the signal model of the microphone array is also studied,including the near-field model,the far-field model and the geometric model of the five-microphone array.An improved scheme based on hidden Markov model keyword recognition algorithm for speech switch array is designed and analyzed.In the endpoint detection phase,endpoints are extracted based on syllables,and these starting points are saved as an array.In the stage of parameter extraction,template training and template matching,the reference template and the Viterbi algorithm are used to match the MFCC parameters of the n syllables,and it is judged whether it is a keyword by comparison with the threshold.Finally,a series of comparative experiments were carried out to influence the placement of the microphone and the distance of the speech signal source on the results,and the simulation experiment was carried out under Matlab.