Running is one of the most basic forms of human movement. For prosthetic wearers, running is helpful to restore the body's motor function and can effectively avoid muscle atrophy of the stump. In recent years, the development of dynamic lower limb prostheses makes it possible for wearers to run, so the research on running intention recognition of prosthetic wearers is very important. In foreign countries, the performance of intelligent prosthesis has been continuously improved, and the functions of corresponding products and prototypes have been gradually improved. Various prosthetic products have their own characteristics in motion state recognition, human-computer interaction, energy efficiency and other aspects. YOBAND has been committed to the research of prosthetic control methods, and designed a variety of dynamic lower limb running motion controllers, which can realize the conversion between running and walking. This paper focuses on gait recognition and control of lower limb prosthesis in different road conditions.

The running state of the lower limb prosthesis wearers is very different from the walking state. Causes the running process to appear stronger oscillations phenomenon, so that the sensor installed on the prosthesis and stump of the signal vibration problem. In order to ensure the reliability of lower limb movement information, single feature of movement information can not meet the requirements. In combination with the study found that on the basis of the residual limb end rectus place ham and knee joint parts is the different movement modes is a typical signal, so multiple electrical sensors, accelerometers installed in strands of rectus muscle and knee, thigh residual limb end acceleration information collected, electrical information and prosthetic knee joint acceleration information. The multi-dimensional feature vectors were constructed by combining the plantar pressure information of both lower limbs as the input of running gait recognition model. In addition, many characteristic parameters of dealing with problems related to the recognition rate and stability of the classifier and SVM on multiple input characteristics of the fault detection problem has a good classification effect, and has been employed in the fall in the prosthetic abnormal gait recognition, so the selection of SVM as running gait classifier, and classification model parameters were optimized by PSO to solve the global optimization problem, Improve the accuracy of SVM order classification model.

First, gait analysis and data collection

Before recognizing the running gait of lower limb prosthesis, the running motion pattern of human lower limb is analyzed. Running and walking of lower limbs are movements of human body alternately, but there are fundamental differences in movement structure. In the movement process of lower limb walking, there is no phenomenon of flying, and there is always one leg in contact with the ground. The movement characteristics of walking can be described as one foot support and two feet support each other alternately. In the process of running, the lower limbs will appear in the air after one foot is lifted off the ground at the end of the support. Then the other leg falls to the ground, and there is a completely different movement process of supporting and flying with one foot in turn.

The running gait recognition of lower limb prosthesis takes the prosthesis side as the research object, and discusses the single step running cycle dominated by the prosthesis side. Therefore, the support stage of the healthy limb side is divided into the mid-stage of the prosthesis side in the air, and the whole running cycle is divided into the support stage and the air stage. The aerial stage is divided into three stages, which are the early stage, the middle stage and the late stage of the aerial. The body performance is the aerial folding, the supporting swing leg and the aerial forward extension. By analyzing the characteristics of human running, it is found that the state of lower limb support is related to running speed during running. When running at low pace, the support phase is heel first; When the walking speed reaches a certain level, the support period changes to the forefoot first landing, but this critical value is affected by many factors such as height and age of different groups, and there is no strict standard. In the actual research, the communication with prosthesis wearers and rehabilitation therapists found that there was a certain difference between running coordination of lower limb prosthesis wearers and normal people.