Acoustic Seabed Classification using Fractional Fourier Transform and Time-Frequency Transform Techniques

In this paper we present an approach for processing sonar signals with the ultimate goal of ocean bottom sediment classification. Work reported is based on sonar data collected by the Volume Search Sonar (VSS), one of the five sonar systems in the AN/AQS-20. Our technique is based on the Fractional Fourier Transform (FrFT), a time-frequency analysis tool which has become attractive in signal processing. Because FrFT uses linear chirps as basis functions, this approach is better suited for sonar applications. The magnitude of the bottom impulse response is given by the magnitude of the Fractional Fourier transform for optimal order applied to the bottom return signal. Joint time-frequency representations of the signal offer the possibility to determine the time-frequency configuration of the signal as its characteristic features for classification purposes. The classification is based on singular value decomposition of the Choi William distribution applied to the impulse response obtained using Fractional Fourier transform. The set of the singular values represents the desired feature vectors that describe the properties of the signal. The singular value spectrum has a high data reduction potential. It encodes the following signal features: time-bandwidth product, frequency versus time dependence, number of signal components and their spacing. The spectrum is invariant to shifts of the signal in time and frequency and is well suited for pattern recognition and classification tasks. The most relevant features (singular values) have been mapped in a reduced dimension space where an unsupervised classification has been employed for acoustic seabed sediment classification. The theoretical method is addressed and then tested on field sonar data. In our classification we used the central beams. Good agreement between the experimental results and the ground truth is shown. A performance comparison of our method to a k-means based technique is also given. Results and recomme- - ndations for future work are presented