EFFICIENT CLUSTER IDENTIFICATION FOR MEASURED ULTRA-WIDEBAND CHANNEL IMPULSE RESPONSE IN VEHICLE CABIN

Although the automatic and robust cluster identification is crucial for ultra-wideband propagation modeling, the existing schemes may either require interactions with analyst, or fail to produce reasonable clustering results in more universal propagation environments. In this article, we suggest a novel cluster identification algorithm. Rather than assuming the limited exponential power decay characteristics on UWB channels, from a novel perspective cluster identification is formulated as the local discontinuity detection based on wavelet analysis. Firstly, in order to comprehensively reflect the prevailing amplitude changes induced by new clusters, the moving averaging ratio is extracted from the measured UWB channel impulse responses. Subsequently, the appealing local-transient analysis ability of wavelet transform is properly exploited, and a computationally efficient cluster extraction method is developed. Distinguished from the subjective visual inspection and excluding any analyst interaction, the presented scheme can automatically discover multiple clusters. Our algorithm is premised on the general amplitude discontinuity, and hence is applicable to various complicated operation environments. Moreover, the produced clustering results, essentially depicting realistic physical propagations, are also independent of parameter configurations. Experiments on both simulated channels and the measured data in typical vehicle cabin further validate the proposed method.