Spiking Neuron Signal Encoding via Sparse Temporal Representation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for encoding and decoding continuous time signals using spike trains are not energy efficient and lack an effective end-to-end signal processing framework for deterministic representation.
Innovation Solution
A method and system for signal coding and reconstruction based on spiking neuron modeling, involving the generation of spike train representations, determination of reconstruction coefficients, and conversion of input signals into spike trains using convolution kernels with time-varying thresholds, followed by decoding to reconstruct the original signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If rate code encoding is used to represent continuous time signals using spike trains, then the signal representation is achieved, but the energy efficiency is poor and the spike trains are not sparse
Solution Approach 1:
The patent transforms the encoding approach by changing from rate-based parameters to temporal-sparse parameters. Spike trains are generated with sparse temporal patterns rather than continuous rate coding, fundamentally altering how signal information is represented in the neural spike domain to achieve both sparsity and energy efficiency
Solution Approach 2:
The patent extracts only the essential signal features that need to be represented, rather than encoding the full continuous signal. By identifying and encoding only critical signal characteristics through sparse spike patterns, the system achieves efficient representation with reduced spike train density
2Use of energy by moving object
If sparse spike trains are used to improve energy efficiency, then energy consumption is reduced, but an effective end-to-end signal processing framework is lacking
Solution Approach 1:
The patent creates a universal end-to-end framework that handles multiple signal processing functions (encoding, sparse representation, decoding, reconstruction) within a unified system. This multi-functional approach ensures that sparse spike trains can be effectively processed through complete signal pipelines without requiring separate specialized components for each function
Solution Approach 2:
The patent introduces convolution kernels with time-varying thresholds as intermediary components that bridge the gap between sparse spike generation and signal reconstruction. These kernels act as mediators that transform sparse temporal patterns into meaningful signal representations while maintaining the energy efficiency benefits of sparsity
3Quantity of substance
If convolution kernels with time-varying thresholds are used for encoding, then sparse spike train generation is achieved, but the encoding and decoding process complexity increases
Solution Approach 1:
The patent performs preliminary setup of convolution kernels with pre-defined time-varying threshold characteristics before the actual encoding process. By preparing the encoding framework in advance with optimized kernel parameters and threshold functions, the system reduces the computational complexity during real-time sparse spike generation while maintaining high sparsity
Data Source
AI summary
A method and system are directed to signal coding and reconstruction. The method comprises receiving an input signal, generating a spike train representation based on the input signal, determining a plurality of reconstruction coefficients based on the spike train representation, and generating a reconstructed signal based on the plurality of reconstruction coefficients.