Anomaly Detection From the Signal of Low-Cost Laser Device Without the False Alarm and the Missing

Abstract

We present a low-cost intruder detection system that recognizes an anomaly using a parametric statistical technique, of which the complexity of computation is linear for data size and the probability of false alarm and miss are respectively zeros. To make the perfect intruder detection system inexpensive and effective, we adopt the Internet of Things (IoT) technologies. The IoT devices work with limited power, a little memory, and small computational power. However, the low-cost controllers, such as Arduino are capable of computing the fast Fourier transform. As the test statistics for discriminating whether the Line-of-sight is lost or not, we propose a signal-to-noise ratio (SNR) and also use the power of the matched-filtered signal. By using the coefficients of the fast Fourier transform of the sampled signal, we jointly analyze the signal in the time and frequency domains. If we use the proposed SNR as the test statistics for detection, and if the sample size is greater than or equal to 256, the false alarm probability and the miss probability of the proposed detector are zeros. The experimental results show that the maximum SNR of the case with disturbance is lower than the minimum SNR of the case without disturbance by 14.49 dB.