Design Of Embedded Speech Recognition System For Automotive Electronic Control

It is always the wish of many researchers to let the machine understand human language.With the continuous progress of speech signal processing technology and the development of integrated circuit technology,it is already possible for the machine to understand human language and make actions.At present,voice recognition technology gradual marketization,and more and more people are beginning to notice the unique advantages of voice in the field of device control.In the past,the speech recognition system mainly operated on PCs and dedicated servers.Because these devices are not easy to carry and cost,it is necessary to transplant the speech recognition system to a more portable embedded system.This article has developed based on this background.Research on Embedded Speech Recognition System for automotive control.This project is based on a detailed understanding and research on the status quo of speech recognition at home and abroad.Based on the requirements of automotive electronic control,the objectives and tasks of embedded speech recognition systems are proposed,including hardware system construction,speech recognition algorithm simulation and software system.The hardware system takes the ARM microprocessor as the core,and the processor periphery includes the speech signal acquisition module,the data storage module,and the result display module.These modules are interconnected to form a hardware system that can support the entire speech recognition process.The algorithm firstly elaborated the connection and difference of the speech recognition algorithm on the PC and in the embedded device,explained the necessity of simulating the speech recognition algorithm on the PC,and used the Matlab tool to preprocess the speech signal and detect the endpoints.And feature extraction algorithm for simulation.In the aspect of feature extraction,the advantages and disadvantages of linear prediction coefficient(LPC)and MFCC were compared,and MFCC with higher recognition success rate was selected as the speech recognition feature parameter.In terms of recognition algorithms,the effect of the dynamic time planning(DTW)algorithm is then verified for the problem of unaligned feature vectors of speech signals in speech recognition.The HMM training and recognition algorithms are simulated,and a scheme to solve the refusal recognition problem in isolated speech recognition is proposed.Finally,by comparing the advantages and disadvantages of the above two recognition algorithms,the feasibility of the isolated word speech recognition system scheme is verified on the PC,and it is determined that the isolated word speech recognition based on the HMM is selected as the final scheme.The voice recognition algorithm was transplanted into the embedded system to realize the voice control of the car lights,and the rejection rate of recognition success rate and interference words was tested.The test results show that when the distance between the person and the microphone is within 1 meter,the recognition success rate of isolated words is close to 90%,and the rejection rate exceeds 70%.