Single Acoustic Sensor-Based Time–Frequency Spectrum Sensing Approach for Land Vehicle Detection

Abstract

This paper introduces a time-frequency spectrum sensing (TFSS) approach based on the short-time Fourier transform (STFT) for vehicle detection using a single acoustic sensor. Spectrum sensing on time-frequency plane offers various applications including vehicle detection, tracking, and classification. The proposed technique estimates the threshold from the power spectral density (PSD) distribution of white Gaussian noise (WGN) which is fitted with Chi-square distribution. This threshold is further used to extract the dominating frequency harmonics and detect the presence of a vehicle from the PSD of the signal on the spectrogram plane. Various performance measures are computed on experimentally generated datasets of the moving vehicles for evaluation of the classical amplitude-based energy detector (ED) and proposed algorithm. The main challenging issue for vehicle monitoring system is, to detect the moving vehicle with a minimum probability of false alarm rate at a low signal-to-noise ratio (SNR). The TFSS can detect the signal at a low SNR level up to −11dB with reduced false alarm rate by 50% as compared to ED. It also provides satisfactory performance measures and receiver operating characteristic curve.