Peak Detection in Noisy Sequential Data Using Decayed Threshold Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying periodic signals in noisy environments is challenging due to unpredictable noise and spikes, as traditional methods require stationary frequency and amplitude assumptions that may not hold for non-stationary signals.
Innovation Solution
A method that uses a probability weighting function derived from peak-to-peak intervals to enhance peak detection accuracy, involving a decayed threshold function and recursive peak correction, which partitions data into segments and weights surrounding points to identify local maxima, improving peak detection in noisy signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional filter or time-frequency domain procedures are used to identify periodic signals, then identification can be performed when signal properties are known, but the method fails when signals have unpredictable properties such as non-stationary period or arbitrary morphology
Solution Approach 1:
The patent implements a dynamic peak identification method that adapts to non-stationary signals by using a sliding window approach and dynamically adjusting the search range for peaks. The algorithm continuously updates its parameters based on the local signal characteristics within each window, allowing it to handle varying periods and morphologies while maintaining reliable peak detection through local adaptation rather than global assumptions.
Solution Approach 2:
The patent changes key parameters dynamically during processing: the search range for peaks is adjusted based on the local period estimate, the window size is adapted to signal characteristics, and the threshold for peak identification is modified according to local noise levels. These parameter changes enable the algorithm to maintain reliability across signals with arbitrary morphology and non-stationary properties.
2Measurement precision
If probability weighting function is applied to enhance peak detection, then peak identification accuracy is improved, but computational complexity increases due to recursive processing
Solution Approach 1:
The patent applies preliminary action by first estimating the local period and creating an initial probability weighting function based on historical peak information before performing the actual peak detection. This preliminary setup allows the recursive refinement process to start from a informed position, reducing the number of iterations needed and thus lowering computational complexity while maintaining high accuracy.
Solution Approach 2:
The patent implements partial action by limiting the recursive refinement process to a maximum number of iterations and by applying the probability weighting function only in regions where peaks are likely to occur (within the dynamically determined search range). This partial application reduces unnecessary computations while still achieving high peak detection accuracy in the critical regions.
Data Source
AI summary
A method implemented through an electronic system for processing a sequential data to identify possible peak points is disclosed. The method defines a decayed threshold function and partition the sequential data into a plurality of segments by grouping each data point with surrounding data points into one of the segments. After that, a plurality of weighted segments are derived through weighting the surrounding data points by the decayed threshold function in each of the segments, and the peak points are identified through corresponding weighted segment.