Dynamic Vision Sensor Noise Filtering for Faster Event Timestamping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dynamic vision sensors (DVS) experience delays in event timestamping due to noise events from thermal leakage and parasitic photocurrents, which affect temporal resolution, particularly in high-speed imaging applications.
Innovation Solution
A noise filtering circuit that filters out noise events by identifying spatially and temporally correlated events, using a timeframe to distinguish between actual and noise events, thereby reducing resource consumption and delays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all events are processed without filtering, then no events are lost, but processing delays increase and temporal resolution deteriorates
Solution Approach 1:
The patent extracts and removes noise events from the event stream before processing. The noise filtering circuit identifies and filters out events that are spatially and temporally correlated with previously detected events, separating useful signal events from noise events. This extraction principle directly resolves the contradiction by removing problematic events that would otherwise cause processing delays while preserving temporal resolution for genuine events.
Solution Approach 2:
The patent applies preliminary filtering action to events before they enter the main processing pipeline. The noise filtering circuit performs preliminary discrimination of events based on spatial and temporal correlation criteria, pre-processing the event stream to eliminate noise events ahead of time. This preliminary action prevents noise events from consuming processing resources and causing delays in the main timestamping and processing stages.
2Productivity
If noise filtering is applied, then processing delays are reduced, but device complexity increases
Solution Approach 1:
The patent segments the pixel array into multiple groups, with each group having its own independent noise filtering circuit. This segmentation allows parallel processing of events from different pixel groups, improving throughput while keeping each individual filtering circuit relatively simple. The segmentation principle resolves the contradiction by distributing the filtering workload across multiple simpler units rather than requiring one complex centralized filter.
Solution Approach 2:
The patent applies partial filtering by focusing only on spatial and temporal correlation criteria rather than implementing comprehensive event analysis. The filtering circuit performs a limited but effective subset of event validation, checking only the necessary spatial and temporal parameters to identify noise events. This partial action achieves sufficient noise reduction without requiring overly complex filtering logic.
3Use of energy by moving object
If noise events are filtered out, then resource consumption is reduced, but measurement precision may be affected
Solution Approach 1:
The patent implements feedback mechanisms where the noise filtering circuit continuously learns from detected events and adjusts its filtering criteria. The circuit uses feedback from previously detected events to refine its understanding of noise patterns versus genuine events, improving filtering accuracy over time. This feedback principle resolves the contradiction by making the filtering process adaptive and self-improving, maintaining high detection accuracy while efficiently reducing resource consumption through intelligent event selection.
Data Source
AI summary
Embodiments of the present disclosure provide a dynamic vision sensor, a method, and a noise filtering circuit for a dynamic vision sensor (DVS). The noise filtering circuit is configured to receive a first request signal in response to a first event detected by one of a group of pixels of the DVS. Also, the noise filtering circuit is configured to trigger a timeframe in response to receiving the first request signal. Further, the noise filtering circuit is configured to receive a second request signal in response to a subsequent second event detected by one of the group of pixels. Also, the noise filtering circuit is configured to forward the second request signal to an arbitration logic if the second event is detected within the timeframe, and to block the second request signal from being forwarded to the arbitration logic if the second event is detected outside the timeframe.