Techniques on ship radiated noise power spectrum feature extraction and realization

Feature extraction of ship radiated noise is the key technique for underwater passsive detection system to recognize and classify the targets. Aiming for this, this paper studies the technique on the real-time feature extraction of continuous and line spectrums and engineering realization based on the power spectrum of radiated noise.The technique, on the basis of segmented fitting algorithm, firstly extracts the continious spectrum features and then subtracts the continious spectrum from the power spectrum, and extracts the line spectrum features from the remainder spectrum.Further more, the paper applies the technique to a real feature extraction system, and details the system´s software and hardware design and realization using digital signal processor(DSP). Experimentation results show that the whole designed system performs well and matches the factors which the new generation underwater intelligent passive detection system expects.