Abstract:
In any passive sonar system the FFT processed data has a wide dynamic range. Extracting the desired signal embedded in noise requires special processing before the data is fed to the sonar displays. In implementing all such techniques it is tried that the data contents are preserved while the noise is reduced as much as possible. This process of flattening, generally called whitening, the noise is called normalization. Normalization is applied to broad band and as well as to the narrowband sonar data. In case of passive sonar systems normalization is applied to the power spectrum versus frequency and bearing parameters. For active sonar systems the returned power is also normalized with range and Doppler. The normalization process is different to compression, mostly used in the display systems, wherein there is greater signal loss in the latter. Therefore, in normalization all efforts are made to preserve the signal to noise ratio. There are many normalization techniques which are in practice in sonar systems design. Some of the prominent techniques are Two Pass Means (TPM), Split Two Pass Mean (S2PM), Split Three Pass Mean (S3PM), Split Average Exclude Average (SAXA) and Order Truncate Average (OTA). All these techniques have certain merits and demerits with regards to different applications and available real time processing resources. We have examined these techniques in detail for passive sonar application and have experimented with simulated and real sonar data. The sonar data used for this evaluation constitutes broadband and narrowband spectrum. The data is collected over platform speed from 1 knot to 14 knots. Therefore, sufficient noise and signal strength variation was available to validate the simulated results. In the paper we first present the significance of normalization in sonar systems in general and in passive sonar systems in particular. This is followed by detailed analysis of the aforementioned techniques. Results obtained by applying simulated data are then presented. This is followed by comparison of the aforementioned techniques and their shortcomings. On the basis of these shortcomings the improved normalization model and algorithm was developed.
Page(s):
125-133
DOI:
DOI not available
Published:
Journal: Proceedings 5th International Bhurban Conference on Applied Sciences and Technology , Volume: 0, Issue: 0, Year: 2007