Event-Driven Temporal Convolution for Asynchronous Pulse Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing signal processing techniques, such as uniform sampling, are inefficient for signals with reduced frequency content or periods of silence, leading to excessive power consumption, particularly in energy-scarce applications, and are computationally complex for classification tasks in machine learning systems.
Innovation Solution
The implementation of event-driven temporal convolution for asynchronous pulse-modulated sampled signals, which processes signals only when events occur, using a convolutional output calculated directly from event-driven input samples and interpolating between events, allowing for energy-efficient neuromorphic systems and reducing computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If uniform sampling is used for signal processing, then signal reconstruction accuracy is maintained, but energy consumption increases and computational complexity increases
Solution Approach 1:
The patent implements dynamic sampling where the sampling rate is not fixed but adapts based on signal activity. The system processes signals only when events occur (changes in signal values), rather than using uniform continuous sampling. This dynamic approach maintains reconstruction accuracy for active signals while consuming minimal energy during static periods, directly resolving the contradiction between accuracy and energy consumption.
Solution Approach 2:
The system uses event-driven periodic processing where computation is triggered by signal events rather than continuous periodic sampling. The convolution operation is performed only when relevant events occur in the input signal, creating an irregular periodic action pattern that reduces overall computational load and energy consumption while preserving necessary signal information for accurate reconstruction.
2Measurement precision
If uniform sampling is used for signal processing, then signal reconstruction accuracy is maintained, but computational complexity increases
Solution Approach 1:
The patent extracts and processes only the essential event-driven components of the signal rather than processing the entire continuous signal stream. By taking out only the relevant events (changes in signal values) and performing convolution operations solely on these extracted events, the system maintains reconstruction accuracy while dramatically reducing computational complexity compared to processing all uniformly sampled points.
Solution Approach 2:
The computational process is made dynamic, adjusting the amount of computation based on signal activity. During static periods with no events, computational operations are minimized or suspended. When events occur, the system dynamically performs necessary convolution operations. This dynamic computational approach maintains accuracy when needed while reducing overall computational complexity.
3Measurement precision
If high sampling rates are used, then signal processing accuracy is improved, but energy consumption increases
Solution Approach 1:
The system implements dynamic sampling rate adjustment where the effective sampling rate varies based on signal activity. During periods of high signal activity with frequent events, the sampling rate increases to capture necessary details for accurate processing. During static periods with no events, the sampling rate effectively drops to zero, eliminating unnecessary energy consumption while preserving accuracy for active signal processing.
Solution Approach 2:
The patent changes the sampling parameter from a fixed high rate to a variable rate determined by event occurrence. The sampling interval dynamically adjusts based on signal characteristics and event density, allowing the system to maintain high effective sampling rates only when necessary for accuracy while using lower or zero sampling rates during periods where high accuracy is not required, thus optimizing energy consumption.
Data Source
AI summary
A method of processing asynchronous event-driven input samples of a continuous time signal, includes calculating a convolutional output directly from the event-driven input samples. The convolutional output is based on an asynchronous pulse modulated (APM) encoding pulse. The method further includes interpolating output between events.