Interference excision algorithm for Frequency Hopping Spread Spectrum based on Undecimated Wavelet Packet Transform

This paper presents an algorithm for reducing narrowband interference effects in Frequency-Hopping Spread-Spectrum (FH-SS). The method is based on the Undecimated Wavelet Packet Transform (UWPT), and it improves the performance obtained in previous methods. Some experimental results illustrate the suitability of the algorithm. Introduction: Frequency Hopping Spread Spectrum is a digital modulation, which is robust against noise and interference, and generally used in noncoherent systems. Despite of this robustness, FH-SS systems can suffer from narrowband high energy interference [1]. Different techniques have been developed to eliminate interference in SS communications. These techniques are based on the comparison among signal and interferences, using adaptive filters and transform domain filtering [2]. This paper focuses on FH-SS systems with transform domain filtering, using wavelet transform (WT). WT analysis has special characteristics to confine and manage the signal spectrum in subbands. The algorithm discards the contaminated or jammed subbands at the synthesis stage. However, an interband spectral leakage that cannot be solved with fixed time-frequency resolutions can appear during this process. Thus, in this paper, irregular time-frequency decomposition by Wavelet Packet Transform (WPT) [3] is proposed. WPT allows selecting the subband tree structure filters [3]. This paper proposes a new algorithm based on the UWPT that uses frequency shifts to confine the narrowband interference within one transform subband. FH-SS receiver with interference excision: The FH-SS receiver, illustrated in Fig. 1, is a standard FH-SS receiver with one additional block for interference excision. Input signal r(t) is demodulated in quadrature for all the FH-SS frequencies fc, where c∈[0,...,m-1], and m is the total number of frequencies. Each demodulator output is low pass filtered, sampled and finally divided and processed in blocks by the excision filter. These blocks are: Xcr+jXci, c∈[0, ..., m-1]. After the excision filter, the signal is free from interferences and each symbol is detected with a non-coherent energy threshold. A similar scheme is proposed for DS-SS systems in [4]. The excision filter: The UWPT is a translation invariant and redundant transform, where no decimation is done after the analysis filtering [5]. Thus, analysis and synthesis filters are different at each level, since zeros need to be added in-between the filters’ coefficients. At the “i” level, 2-1 zeros are introduced in-between each pair of coefficients. As shown in Fig. 2, the UWPT at every resolution level divides the spectrum in two, due to its scaling property [6]. In this case, 16 coefficients Daubechies filters [6] have been used. The algorithm implements an analysis process, running from the level with more frequency resolution (level “J”), to the one with less resolution (level “1”). At this first level “J”, the input vector is the original signal from a pair of demodulators XJ=(XJr+jXJi). The algorithm shifts the signal in the frequency domain, to center the interference in a subband. The filtering process is done for every frequency shift, d(J). The number of shifts is given by the bandwidth of the filters (BJ), d(J)∈[0,..,BJ]. Thus, the output vectors using a periodic extension of the input signal are: ) 1 ( )) 2 ( ( ) h( ) ( 1 0 1 ) ( , ) ( , 0 , ∑ −