A Novel Acoustic Characteristic Extraction Algorithm for Traffic Volume Estimation

Traffic volume information is widely used in all aspects of Intelligent Transportation Systems (ITS), such as transportation planning, traffic states identification, traffic management, safety analysis, and so on. In recent years, acoustic sensors are gradually applied to the detection of various traffic parameters. In this paper, acoustic data sets acquired from acoustic sensors are utilized to estimate the road traffic volume. The short-term energy (STE) algorithm and the energy to zero crossing rate (EZCR) algorithm are usually applied to acoustic analysis, and they both perform well under some simple circumstances, however, some urgent problems remain unresolved under certain complex conditions. One of such issues occurs when adjacent vehicle-pass signals (VPSs) intersect partially, seriously decreasing the accuracy of endpoint detection of VPSs, hampering the algorithm ability to maintain satisfactory traffic volume estimation. Another difficulty arises while multiple lanes are considered: some special VPSs cannot be detected. To solve these problems, a novel acoustic characteristic is defined, and an acoustic-based characteristic extraction algorithm for traffic volume estimation, entitled triangular wave analysis (TWA), is proposed. Comparing the TWA algorithm with the STE and EZCR algorithms, experimental results demonstrate the viability of the proposed algorithm in the case of intersectant VPSs.