Cascaded Feature Detection for Low-Power Stream Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data stream analysis systems face high computational and hardware costs due to inefficient processing during periods of low interest, leading to high false positive rates and suboptimal power consumption.
Innovation Solution
Implementing a cascade of algorithms with increasing computational complexity, using a first simpler algorithm to identify and buffer 'interesting' data portions, followed by a more complex algorithm for detailed analysis, while dynamically adjusting thresholds to retain only a fixed proportion of the data stream.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a complex algorithm is used to fully process every frame, then the understanding of what is of interest is improved, but the computational power and hardware cost increase significantly
Solution Approach 1:
The processing system is segmented into multiple stages: a first algorithm processes all frames to identify candidate regions, a buffer stores selected portions, and a second algorithm processes only the buffered portions. This segmentation allows different processing depths for different data portions, reducing overall computational power while maintaining detection accuracy for interesting content.
Solution Approach 2:
Different processing qualities are applied to different portions of the data stream. Frames identified as containing interesting content receive full processing with the second algorithm, while other frames receive only the first algorithm's processing. This local quality approach maintains high detection accuracy for relevant content while reducing average computational power consumption.
2Use of energy by moving object
If a motion detector is used to filter frames before processing, then the power consumption is reduced, but the false positive rate increases
Solution Approach 1:
A buffer acts as an intermediary between the first algorithm and the second algorithm. The buffer stores portions of the data stream identified by the first algorithm, allowing the second algorithm to process these portions with reduced computational requirements while maintaining reliable detection. This intermediary structure reduces false positives compared to simple motion detection.
3Reliability
If the hardware is designed to handle peak data rates, then the reliability during peak operation is improved, but the hardware cost increases
Solution Approach 1:
The system dynamically adjusts the amount of data sent to the second algorithm based on the buffer contents and processing needs. During peak operation, the buffer manages data flow to ensure the second algorithm receives appropriate input without overwhelming the hardware. This dynamic approach allows hardware to be optimized for average load while maintaining reliability during peaks.
Data Source
AI summary
A computer implemented method of data processing comprises receiving an input data stream; identifying, by a first algorithm, one or more first content features of the data stream; determining, based on at least one output of the first algorithm, one or more portions of the data stream to be stored in a first buffer; storing the determined portions in the first buffer; identifying, by a second algorithm, one or more second content features of the data stream using the stored portions of the data stream; and generating, for output to a user, an output based on the identified second content features; wherein the second algorithm is computationally more complex than the first algorithm.