| With the development of artificial intelligence,smart life has affected today's lifestyle.Traditional human-computer interaction has gradually failed to meet people's needs.Natural human-computer interaction technology has become a research hotspot at the moment.As an important expression,gestures have become a breakthrough in human-computer interaction technology.Gesture recognition based on active ultrasound has become a new research area due to its low cost,low power consumption,and is less susceptible to external interference.In this thesis,the hardware for gesture recognition based on active ultrasound technology is designed,and the algorithm of gesture recognition is discussed and good results are obtained.The main work of this thesis is as follows:(1)A set of low-cost,low-power hardware devices based on active ultrasound technology was designed.Using the STM32F407VGT6 as the core,single-emission and four-receiving methods are used,and low-cost transducers and MEMS microphones with stable frequency response curves in the ultrasonic range are used for ultrasonic transmission and reception.On this basis,the design of data storage,data transmission and other procedures were carried out,and the gesture recognition algorithm was transplanted.(2)An improved zero-crossing detection is proposed to estimate the frequency of the gesture signal,and the gesture is intercepted by the processing of the amplitude domain.For the data collected in the experiment,the use of improved zero-crossing detection method to estimate the frequency,while achieving good results,greatly reducing the time and resource consumption.Using the method of combining average voltage and low-pass filtering,the amplitude information is processed.FBGD method is used to intercept the gesture through the amplitude domain feature,and good results are obtained.(3)The gestures were classified using a combination of support vector machine and BP neural network.In order to facilitate the algorithm migration,a two-step classification algorithm was used.The gestures were divided into vertical gestures and horizontal gestures using a support vector machine.Then use BP neural network to calculate in each group.Finally,the recognition and classification of the seven gestures of up,down,left,right,push,pull,and push and pull were completed,and good results were achieved. |
