Sign Language Recognition Research Based On EMG And Six-axis Gyroscope Sensors

The research of sign language recognition is of great significance for improving the communication between deaf-mute and normal people and helping them to better integrate into society.Sign language recognition has a variety of data collection forms.Among them,the multi-sensor-based sign language recognition method has received more and more attention because it is more convenient to wear than data gloves and has less interference than video recognition background.Surface Electromyography signal(s EMG)is often used in sign language recognition research because of its ability to reflect subtle movements such as finger flexion and extension and convenient collection.Acceleration signal has also become sign language in recent years because it can reflect information about arm swing Identify research hotspots.Combining the two types of sensors to complement each other can capture more complete sign language data information,thereby improving the recognition accuracy.Aiming at the problem of low accuracy of sign language recognition,this paper proposes a hybrid particle swarm algorithm Support vector machine(Hybrid Particle swarm algorithm Support vector machine,HPSO-SVM)based on the sign language data collected by the electromyography sensor and the six-axis gyroscope sensor.Chinese sign language for recognition.The main work is as follows:(1)Signal collection: A more practical and effective sensor combination device is designed to collect sign language data.The characteristics of the EMG signal produced by the arm muscles are studied,and the muscle area with active signal when performing sign language actions is selected as the collection area,which provides a guarantee for a good data source.(2)Signal effective activity segment extraction: Aiming at the difficult extraction of complex motion activity segments,an improved algorithm of short-time energy method is proposed.By adding an adaptive fault-tolerant length,it is judged whether the following segment of data has ended or is a small segment of the complex motion.End,so as to effectively extract the EMG signal activity segment of all actions.(3)Feature extraction: according to the different characteristics of the EMG signal and acceleration and angular velocity signals,different features are extracted respectively.The features are clustered and analyzed,so that the features with better classification effect are selected,and the complexity of the algorithm operation is reduced.(4)Sign language classification method: The HPSO-SVM algorithm is proposed on the basis of the SVM algorithm.The parameter penalty factor c and the kernel function parameter g of the SVM are used as the coordinates of the particles in the particle swarm algorithm,and the current particle is added to the global optimal solution.The local optimal solution with a center and a radius of R reduces the risk of particle swarms falling into the local optimal solution,so that the hybrid particle swarm algorithm can solve the optimal parameters of the support vector machine and improve the accuracy of sign language recognition.At the same time,it is compared and analyzed with SVM,BPNN algorithm and DT algorithm.It is concluded that the recognition rate of the HPSO-SVM algorithm proposed in this paper is slightly better than other algorithms,and the recognition accuracy rate reaches 96.67%.Reflects the superiority of this algorithm.