Real Time Pattern Matching with Dynamic Normalization
This addresses the problem of robust real-time pattern matching for applications dealing with distorted time series data, though it appears incremental as it builds on existing Dynamic Time Warping techniques.
The paper tackles the challenge of pattern matching in time series data streams affected by noise, amplitude variations, and time distortions by introducing a dynamic z-normalization mechanism and a Dynamic Time Warping-based method, achieving high operational characteristics compared to state-of-the-art methods.
Pattern matching in time series data streams is considered to be an essential data mining problem that still stays challenging for many practical scenarios. Different factors such as noise, varying amplitude scale or shift, signal stretches or shrinks in time are all leading to performance degradation of many existing pattern matching algorithms. In this paper, we introduce a dynamic z-normalization mechanism allowing for proper signal scaling even under significant time and amplitude distortions. Based on that, we further propose a Dynamic Time Warping-based real-time pattern matching method to recover hidden patterns that can be distorted in both time and amplitude. We evaluate our proposed method on synthetic and real-world scenarios under realistic conditions demonstrating its high operational characteristics comparing to other state-of-the-art pattern matching methods.