Research On Intelligent Wheelchair Control System Combining Voice And EMG

Intelligent wheelchair is based on electric wheelchair using intelligent robot technology.Its research involves many fields,such as machinery,control,sensors and artificial intelligence.At present,because of its good autonomy and practicability,intelligent wheelchair can effectively solve the problem of the elderly and the disabled,which has attracted more and more attention from domestic and foreign researchers.Among them,with the development of artificial intelligence,voice,electromyography and other hand-free wheelchair control methods have also been favored by a large number of researchers.Combining with the research direction of the subject,starting from the background and significance of the topic,this paper analyses the characteristics of voice and EMG signals,studies the process of acquisition and preprocessing,feature extraction and pattern classification of voice and EMG signals respectively.At the same time,a commercial wheelchair is reformed to construct a intelligent wheelchair including motor drive,power circuit,audio acquisition,EMG acquisition,wireless communication and so on.By building the recognition system of voice and EMG based on Lab VIEW platform,further explores the control method of electric wheelchair which integrates voice and EMG information.The research results are as follows:(1)In view of the fact that speech signals are easily disturbed by background noise and acquisition instruments in the process of acquisition,the traditional empirical mode noise reduction method can easily lead to the loss of low-frequency effective speech signals.This paper proposes the method of combining empirical mode decomposition(EMD)and wavelet analysis(Wavelet)to reduce the noise of speech signals.This method can retain low-frequency speech signals and the effective speech segments after noise reduction.The mean square error and signal-to-noise ratio are 0.0358,-8.954 and-7.225 db respectively when the signal-to-noise ratio is-10 db.Because EMD-Wavelet algorithm mainly reduces the interference of high frequency noise to speech signal,in order to compensate for the weakness of EMD-Wavelet algorithm which is insensitive to low frequency noise,Meier cepstrum parameter(MFCC)with good anti-noise performance of low frequency noise will be used as speech recognition feature parameter.Principal Component Analysis(PCA)is used to reduce the dimension of MFCC,and dynamic time warping(DTW)is used to improve the speechrecognition speed by 26.2%.And the accuracy of speech recognition is improved by 6.2%.(2)Because the signal-to-noise ratio of EMG signal is relatively low,the effect of denoising using wavelet threshold is not obvious.A method combining wavelet entropy and empirical mode decomposition is proposed,which reflects the characteristics of noise signal.According to the principle of entropy increase,the noise threshold of each layer of wavelet decomposition coefficient is selected adaptively,which is more accurate than the traditional method of noise threshold selection.When the SNR of the denoised signal is 10 db,the mean square error,SNR and correlation coefficient of the denoised signal are 0.2183,18.4 d B and0.5277 respectively.In order to improve the accuracy and efficiency of electromyographic recognition classifier,particle swarm optimization(PSO)is proposed to optimize the parameters of double support vector machine(TWSVM).The training time of the model is2.75 times longer than that of SVM.As a result of introducing PSO algorithm to optimize the parameters of the model,the time of TWSVM is slightly lower than that of PSO-TWSVM model,but the PSO-TWSVM model is significantly higher than that of TWSVM and SVM,and the classification accuracy is improved by 4.2%.Compared with SVM,PSO-TWSVM has more advantages in recognition accuracy and speed.(3)By completing the hardware and software design of the intelligent wheelchair control system,the experimental platform of the intelligent wheelchair control system integrating voice and EMG was built.Five volunteers were invited to use Lab VIEW voice-EMG control mode to drive the wheelchair to complete the predetermined trajectory.By analyzing the performance of different volunteers in the three control schemes,it was found that the completion time of the path was easy for the operator to be familiar with.Among the three schemes,the accuracy of speech recognition is relatively stable,which fluctuates within the range of 0.4% and the fluctuation range of recognition rate of electromyography is 1.5%.The recognition efficiency of electromyography will directly affect the completion time of the planned path.The completion time of scheme 3 is significantly lower than that of the other two schemes.Therefore,in the actual use of wheelchairs,the operator’s proficiency should be improved as far as possible,and the types of electromyography should be reduced to avoid muscle fatigue.Finally,scheme 3 is chosen as the final control mode of wheelchair: the wheelchair can be switched in different directions by voice command,and a pair of electromyography is used to determine whether the selected direction of motion is executed or not.The research of speech and EMG signal recognition methods is completed in this paper.The recognition accuracy and recognition time are better than the traditional methods.At the same time,the control mode of speech-EMG fusion is more stable than that of single control source,which reduces the misjudgement of unconscious actions,and provides a method ofreplacing the manipulator pole for patients with limb inactivity deficiency,and makes wheelchair operation more simple and humanized.