Signal Processing On Warship Noise And Its Realization By DSP

The analysis of warship noise for identification is an important research topic in underwater acoustic engineering. Traditionally, the warship noise is transformed by FFT into frequency domain for further analyzing and processing. However, frequency resolution is equal to the inverse ratio of the sampling length. Therefore, higher frequency resolution needs longer sampling time, which means more points to be computed in an FFT operation. Thus it is not pragmatic for a real time signal processing system to adopt the frequency-domain method based on FFT.To detect the line spectrum in warship noise, the author gives an approach that hires a bandpass filter to identify the specific line spectrum in the warship noise. The frequency resolution needed in application is turned to the passband bandwidth of the bandpass filter. The narrower bandwidth we design, the higher frequency resolution we obtain.In a practical digital system, the system function after the quantization of coefficients might be quite different from the original one because of the finite word-length effect. Sometimes the originally stable system even turns unstable after quantization. To tackle that problem, the analysis of the error transfer process between the coefficients quantization error and the zero-pole error reveals that the improvement of system structure could reduce the coefficients sensitivity of system function.Finite impulse response filter with an order of several thousands is applicable, thanks to the development of computing capability of DSP. Therefore, within the computing capacity of DSP, instead of infinite impulse response filter, we could use finite impulse response filter with a high order to reduce the impact of coefficients quantization error. The author designs and implements a finite impulse response filter with an order of 5200 on a TMS320VC5502 DSP as an instance for utilizing the strong computing ability of DSP in a real time signal processing system.