Spiking Neural Network Phase Encoding for Cognitive Computing
Lei Zhang
TL;DR
The paper addresses reconstructing arbitrary time series signals using a spiking neural network framework grounded in cognitive informatics. It introduces a DFT-based SNN in which each neuron corresponds to a DFT frequency component and the spiking signal is formed as $s(t)= A e^{i(2\pi f t + \phi)}$ with the network summing $C_k e^{i 2\pi f_k t}$ to realize an inverse DFT. It then develops phase encoding through impulse delays, formalizing the relationship $\Delta\phi= -i\frac{2\pi}{N} n_0$ and demonstrating N=2 and N=4 impulse cases to reveal sequential firing patterns. The work also derives the derivative of the frequency spectrum in polar coordinates, providing a gradient framework for encoding phase and magnitude changes. Collectively, these results offer a biologically plausible, time-aware approach to cognitive signal analysis and real-time, low-power signal reconstruction using SNNs.
Abstract
This paper presents a novel approach for signal reconstruction using Spiking Neural Networks (SNN) based on the principles of Cognitive Informatics and Cognitive Computing. The proposed SNN leverages the Discrete Fourier Transform (DFT) to represent and reconstruct arbitrary time series signals. By employing N spiking neurons, the SNN captures the frequency components of the input signal, with each neuron assigned a unique frequency. The relationship between the magnitude and phase of the spiking neurons and the DFT coefficients is explored, enabling the reconstruction of the original signal. Additionally, the paper discusses the encoding of impulse delays and the phase differences between adjacent frequency components. This research contributes to the field of signal processing and provides insights into the application of SNN for cognitive signal analysis and reconstruction.