Adaptive Window Analysis Device for Temporal Distortion Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data stream analysis methods fail to detect events with temporal distortion and require pre-defined window lengths, making it difficult to identify identical events without prior knowledge and handling non-waveform data with overlapping events.
Innovation Solution
An analysis device that decides a window length, constructs graphs within those windows, forms tensors from multiple graphs, extracts patterns, and identifies residual portions, iteratively adjusting the window length to detect events with temporal distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a fixed window length is used for data stream analysis, then the analysis process is simple and fast, but it cannot detect events with temporal distortion and requires pre-defined window lengths
Solution Approach 1:
The patent applies dynamics by making the window length adjustable and adaptive rather than fixed. The system dynamically determines optimal window lengths based on the characteristics of the data stream and events being analyzed, allowing the analysis window to adapt to different event durations and temporal patterns without requiring pre-defined fixed lengths.
Solution Approach 2:
The patent changes the parameter of window length from a fixed value to a variable that can be optimized. By allowing the window length parameter to be adjusted based on event characteristics and analysis requirements, the system can accommodate events with varying temporal distortions while maintaining analysis effectiveness.
2Measurement precision
If the window length is adjusted to detect events with temporal distortion, then event detection accuracy improves, but the complexity of determining the appropriate window length increases
Solution Approach 1:
The patent applies self-service by enabling the system to automatically determine optimal window lengths without requiring manual configuration or external input. The analysis device autonomously analyzes data stream characteristics and selects appropriate window lengths, eliminating the need for users to understand or configure complex windowing parameters.
Solution Approach 2:
The patent uses feedback mechanisms where the system evaluates the effectiveness of different window lengths based on event detection results and adjusts subsequent window length selections accordingly. This iterative optimization process allows the system to learn from previous analysis outcomes and improve event detection accuracy over time.
3Reliability
If existing analysis methods are used, then simple patterns can be detected, but events that should be identified with identical events cannot be recognized and residual events cannot be effectively analyzed
Solution Approach 1:
The patent applies extraction by separating identified events from the data stream and creating a residual data stream containing undetected or partially detected events. This extracted residual stream is then subjected to further analysis with different parameters, allowing the system to recover additional event information that was missed in the initial analysis pass.
Solution Approach 2:
The patent applies partial action by performing multiple analysis passes on the data stream, each focusing on different aspects or using different window lengths. Rather than attempting to detect all events in a single pass, the system performs successive partial analyses on the residual stream, gradually extracting more event information through iterative refinement.
Data Source
AI summary
An analysis device of the present invention includes: a window length decision part that decides a window length of a data stream including tuples each including a timestamp and an observed value; a graph construction part that constructs a graph representing a preset relation between the tuples within a window obtained by dividing the data steam by the decided window length; a tensor construction part that constructs a tensor or a matrix equivalent to the tensor from a plurality of graphs constructed for a plurality of windows, respectively; a pattern extraction part that extracts a pattern from the tensor or the matrix equivalent to the tensor; and a difference extraction part that extracts a residual portion that does not match the extracted pattern, from the data stream. The window length decision part decides a new window length that is different from the previously decided window length.