Time-frequency transforms, including wavelet and wavelet packet transforms, are generally acknowledged to be useful for studying non-stationary phenomena and, in particular, have been shown or claimed to be of value in the detection and characterization of transient signals. In many applications time-frequency transforms are simply employed as a visual aid to be used for signal display. Although there have been several studies reported in the literature, there is still considerable work to be done investigating the utility of wavelet and wavelet packet time-frequency transforms for automatic transient signal classification. In this paper we contribute to this ongoing investigation by exploring the feasibility of applying the wavelet packet transform to automatic detection and classification of a specific set of transient signals in background noise. In particular, a noncoherent wavelet-packet-based algorithm specific to the detection and classification of underwater acoustic signals generated by snapping shrimp and sperm whale clicks is proposed. We develop a systematic feature extraction process which exploits signal class differences in the wavelet packet transform coefficients. The wavelet-packet-based features obtained by our method for the biologically generated underwater acoustic signals yield excellent classification results when used as input for a neural network and a nearest neighbor rule.

Abstract-Optimal joint detection for interfering (nonorthogonal) users in a multiple access communication system has, in general, a computational complexity that is exponential in the number of users. For this reason, optimal joint detection has been thought to be impractical for large numbers of users. A number of suboptimal low-complexity joint detectors have been proposed for direct sequence spread spectrum user waveforms that have properties suitable for mobile cellular and other systems. There are, however, other systems, such as satellite systems, for which other waveforms may be considered. This paper shows that there are user signature set selections that enable optimal joint detection that is extremely low in complexity. When a hierarchical crosscorrelation structure is imposed on the user waveforms, optimal detection can be achieved with a tree-structured receiver having complexity that is, in typical cases, a low-order-polynomial in the number of users. This is a huge savings over the exponential complexity needed for the optimal detection of general signals.Work in recent literature has shown that a hierarchically structured signal set can achieve oversaturation (more users than dimensions) with no growth in required signal-to-noise ratio. The proposed tree detector achieves low-complexity optimal joint detection even in this oversaturated case.

Time-frequency transforms, including wavelet and wavelet packet transforms, are generally acknowledged to be useful for studying non-stationary phenomena and, in particular, have been shown or claimed to be of value in the detection and characterization of transient signals. In many applications time-frequency transforms are simply employed as a visual aid to be used for signal display. Although there have been several studies reported in the literature, there is still considerable work to be done investigating the utility of wavelet and wavelet packet time-frequency transforms for automatic transient signal classification. In this paper we contribute to this ongoing investigation by exploring the feasibility of applying the wavelet packet transform to automatic detection and classification of a specific set of transient signals in background noise. In particular, a noncoherent wavelet-packet-based algorithm specific to the detection and classification of underwater acoustic signals generated by snapping shrimp and sperm whale clicks is proposed. We develop a systematic feature extraction process which exploits signal class differences in the wavelet packet transform coefficients. The wavelet-packet-based features obtained by our method for the biologically generated underwater acoustic signals yield excellent classification results when used as input for a neural network and a nearest neighbor rule.

No abstract

No abstract