Time-Frequency Signal Processing Based on Fractional Fourier Transform in Passive Sonar Classification.

This paper introduce a new data SONAR classification method based on Short-Time Fractional Fourier Transform (STFrFT) analysis. The passive SONAR system receives the acoustic signals radiated by vessels and attempts to categorize them as a function of the similarities between vessels of the same class.Here, a time-frequency processing and feature extraction method is developed in order to improve the performance of a feedforwardneural network, which is used to classify five classes of vessels.Processing of time-varying signals in fractional fourier domain allows us to estimate the signal with higher concentration than conventional fourier domain, making the technique robust against additive noise, maintaining same computational complexity. With the purpose of dimension reduction and classification improvement, we use Linear Discriminant Analysis (LDA) technique. The feasibility of the proposed technique (STFrFTLDA) has been tested experimentally using a real database. The experimental results show the superiority of the proposed method.