AbstractDoppler sodar is a technology used for acoustic-based remote sensing of the lower planetary boundary layer. Sodars are often used to measure wind profiles; however, they suffer from problems caused by noise (both acoustic and electrical) and echoes from fixed objects, which can bias radial velocity estimates. An experimental bistatic sodar was developed with 64 independent channels. The device enables flexible beamforming; beams can be tilted at the same angle irrelevant of frequency, a limitation in most commercial devices. This paper presents an alternative sodar signal-processing algorithm for wind profiling using a multifrequency stepped-chirp pulse. A noncoherent matched filter was used to analyze returned signals. The noncoherent matched filter combines radial velocity estimates from multiple frequencies into a single optimization. To identify and separate sources of backscatter, noise, and fixed echoes, a stochastic pattern-recognition technique, Gaussian mixture modeling, was used to postprocess the noncoherent matched filter data. This method allowed the identification and separation of different stochastic processes. After identification, noise and fixed echo components were removed and a clean wind profile was produced. This technique was compared with traditional spectrum-based radial velocity estimation methods, and an improvement in the rejection of fixed echo components was demonstrated; this is one of the major limitations of sodar performance when located in complex terrain and urban environments.
Bio-Inspired Covert Active Sonar Strategy
