Grouping Gesture Recognition Based On Hidden Markov Model

The use of wearable devices for gesture recognition is rapidly becomi ng a research focus,which is widely used in behavior detection,sign language recognition and human-computer interaction.Today,with the development of micro electromechanical system(MEMS),it has become possible to produce smaller,lighter sensors and devices that can be worn on people to dete ct human behavior and even smaller limb movements.When actual gesture recognition is performed on a resource-constrained wearable device,the recognition accuracy and the time complexity of the algorithm need to b e considered.At present,many gesture recognition algorithms have been proposed and adopted.Hidden Markov Model(HMM)is the most used gesture recognition algorithm at present.HMM is used in speech recognition for the first time.Because of the similari ty of gesture sequence and speech sequence,HMM is widely used in gesture recogni tion and can achieve a relatively high recognition accuracy.However,due to the high computational complexity of HMM,when it is used on resource-constrained wearable devices to perform gesture recognition,the effect of real-time response can not be achieved,and the user experience is poor and needs to be improved.The computational complexity of the HMM is directly proportional to the size of the data set to be identified,the length of the observed sequence,and the number of states.Reducing the value of these three parameters can reduce the computational complexity,but the recognition accuracy is correspondingly low.Therefore,it is necessary to find a way to reduce the computational complexity of the algorithm and to adapt to the computing power of the mobile terminal while maintaining the recognition accuracy.Aiming at the above problems,this paper proposes a method to reduce the computational complexity of the recognition algorithm by grouping the gestures and to maintain or even improve the recognition ac curacy by setting different HMM for each group.This method consists of three parts: gesture grouping,group model establishment,and the establishment of gesture m odel in each group.Gesture grouping uses K-means++ based methods;group models use table-based methods;gesture models use HMM,and HMM within a group have a similar structure,having different structures with different groups.To validate the effectiveness of the grouping method,12 gestures were defined in this study.These gestures take into account different shapes,orientations,and repetitions,and are representative.And then builds a data acquisition platform,which includes the wearable hardware p latform and the host PC platform,through this platform to collect a large number of gesture data to verify the method proposed in this paper.Experimental results show that the computational complexity of the method proposed in this paper is greatly reduced,and the feasibility of the method is verified,without any loss of recognition accuracy,compared with the standard HMM.