Research On Human Activity Recognition Algorithm Based On HRRP Radar Signal

Human activity recognition is a research hotspot in the field of pattern recognition.It has a wide range of applications in human-computer interaction,security monitoring,medical and military fields.With the development and application of artificial intelligence technology,two main research methods have emerged in the field of human activity recognition:one based on sensor signals,such as various wearable sensors,and one based on visual signals,such as natural images and video clips.Compared with the above two methods,human activity recognition based on radar signals has significant advantages in terms of deployment convenience and environmental robustness,but the recognition algorithm based on two-dimensional radar signals has low noise robustness and poor real-time performance.Therefore,this paper proposes an improved recognition algorithm based on two-dimensional radar signals and a human activity recognition algorithm based on one-dimensional radar signals.Aiming at the problem of noise interference in two-dimensional radar signals,this paper proposes a human activity recognition algorithm based on attention mechanism and convolutional neural network.This algorithm adds an attention mechanism and a spatial pyramid pooling layer to the network structure.Finally,the layered feature map generated by the network is fused to achieve the purpose of using the attention mechanism to reduce noise interference.This paper compares the performance of the proposed algorithm with five common convolutional neural networks under different degrees of noise interference.The results prove that the proposed algorithm can more effectively reduce the impact of noise when the signal-to-noise ratio is higher than-10dB.The recognition accuracy rate of more than 92%can be achieved under the following conditions.Human activity recognition based on two-dimensional radar signals has poor real-time performance.One-dimensional radar signals have lower feature dimensions and better real-time performance than two-dimensional radar signals.In this paper,a human activity recognition algorithm based on a one-dimensional radar signal is proposed.This algorithm uses a bidirectional gated recurrent neural network to process high-resolution range profiles to realize the classification of human activity.This paper compares the performance of the proposed network with recurrent neural networks,bidirectional recurrent neural networks,and unidirectional gated recurrent neural networks.The results prove that this network structure has a good ability to extract features for high-resolution distance profiles.In the test set,the recognition rate reached 97.6%.Finally,it is compared with the recognition algorithm based on deep convolutional neural network and two-dimensional radar signal.The result proves that the human activity recognition algorithm based on a one-dimensional radar signal has higher calculation efficiency and better real-time performance.The required data duration is 5.6%of the deep convolutional neural network.