Abstract
A study of fidelity criteria (i.e., distortion measures) appropriate for data compression of passive sonar signals was conducted. The classical mean squared error (MSE) criterion was found to be a poor choice for this application, as well as other applications wherein the source signal is to be spectrum analyzed. Two new fidelity criteria, the mean incoherence (MIC) and mean log-coherence (MLC), were developed and tested on passive sonar type signals. The experimental results indicated that these criteria provide an accurate assessment of the distortion produced by a data compression system on this type of signal. The rate-distortion functions of a Gaussian source signal, subject either to MIC or MLC fidelity requirements, were then derived. These functions specify the minimum transmission rate (e.g., bits/sample) required in order to reproduce the source signal with a given distortion. Also, the analysis indicates the effect of optimum encoding on the source signal power spectrum. A number of near-optimum encoding schemes were then developed for these new fidelity criteria. [Work supported by Naval Ocean Systems Center.]
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
