Abstract: In nuclear explosion tests, electronic warfare, 6G communication testing, and chip testing, the accurate measurement of ultra-wideband (UWB) complex signals is fundamental for system performance evaluation, physical process reconstruction, and fault target identification, and represents a core capability of broadband time-domain test instruments. However, as electronic signals evolve toward ultra-wide bandwidth, large dynamic range, short-duration bursts, and strong randomness, conventional measurement techniques are constrained by fixed sampling architectures, limited real-time processing capability, and single-domain analysis, leading to three major challenges: 1) broadband sparse signal measurement suffers from the trade-off between bandwidth and resolution, making it difficult to achieve both wide bandwidth and high accuracy; 2) sporadic anomalies in ultra-high-speed data streams are easily overwhelmed by massive normal data, preventing reliable real-time capture; and 3) signal characteristics are distributed across multiple information domains, including the time, frequency, and modulation domains, making conventional single-domain analysis difficultly to achieve synchronized multi-domain measurement. To address these challenges, this paper proposes a reconfigurable hybrid interleaved sampling architecture in the time, frequency, and amplitude domains, together with target feature-driven intelligent sampling, anomaly capture, and multi-domain analysis methods. By autonomously learning the multi-domain feature distribution of signals, the proposed approach dynamically optimizes sampling resources, detection resources, and analysis strategies, enabling high-precision acquisition, lossless capture of sporadic anomalies, and intelligent multi-domain analysis of UWB complex signals. This work advances time-domain test instruments from fixed-configuration passive acquisition toward adaptive acquisition, autonomous perception, and intelligent analysis, providing theoretical foundations and technical support for next-generation intelligent test instruments.