A Special Processor of Passive Infrared Detector Based on Artificial Neural Network

At present, most of the PIR sensors use analog circuits as the detection methods. Their peripheral circuits are complex and time-consuming, so it is difficult to balance the detection accuracy and the complexity of the detection circuit. This paper chooses a programmable processor to solve the above problems. Based on the field programmable gate array (FPGA) with amicrocontroller unit (MCU) soft core, a neural network algorithm is designed to analyze and process the data collected by PIR sensor. The neural network adopted three-layer neural network structure and used the rectified linear unit (ReLU) function and linear function as its activation function. Compared with traditional statistical classification method, this algorithm made greatly improvement in many performance parameters. It provided a theoretical and engineering basis for making a low-power and programmable PIR intelligent control chip, with high commercial value.