Human Recognition Based On Passive Infrared Detector

With more attention paid to the resident environment security, passive infrared detectors (PIR detectors) are widely used in the home security market because of their low-cost and good-concealment. However, PIR detectors are still faced with some underlying issues, such as the high rate of false alarm etc. Compared with the oversea similar products, domestic PIR technology is lack of sufficient and effective signal analysis. There is no effective data-extracting of signal characteristics. The simple methods result in a lower alarm accuracy of the PIR detectors, such as pulse counting/timing and signal amplitude threshold trigger. So it is not conducive to the promotion of related products in practical application.This thesis firstly introduces the characteristics of passive infrared detecting technology, and then gives a deep analysis on the working principle of each component of PIR detectors. Finally, a novel method based on analyzing pyroelectric signals effectively is proposed for designing smart PIR detector. In terms of smart PIR detector, the thesis does exploratory researches as follow:①A new method for human detecting is proposed based on the compound layout of PIR detectors by deeply analyzing the working principle of pyroelectric infrared detectors and optical system. This method is meanful for improving the performance of PIR detector in spatial distribution. The simulation experimental results demonstrate that the compound layout can improve effectively the detecting performance in specific locations.②The thesis proposes to have completed three experiments (human walking by given paths, human moving by random ways, and animal (such as dog) moving by random ways) with self-made PIR detectors. This method makes use of digital oscillograph to collect a great deal of output signals from PIR detector circuit. The signals are the basic for data analyzing.③This thesis analyzes the pyroelectric signals separately in time-domain and frequency-domain and extracts the data characteristics hidden in the signals. And this thesis illustrates the detecting signal distinction characteristics of human moving on spectral power. This thesis extracts human and animal data characteristics by PCA and CCA, classifies the data by NN and SVM, and then compares the recognition rates. ④This thesis designs the circuit of PIR detector for preprocessing the output signals and proposes a novel method for designing smart PIR detector, which can complete human-target smart recognition on the programmable control chip PIC16F877 by extracting the output signal characteristics and using classification algorithms.