Spike-IMU: An Accurate and Low-Power Spiking Neural Network for Pedestrian Velocity Estimation
Junye Zou, Xiaolei Li, Ziyang Meng, Guoqi Li
AI summary
Problem
Deploying accurate pedestrian navigation on low-power edge devices is hindered by the high energy demands of artificial neural networks and the information loss during spike encoding and simplistic neuron dynamics of existing spiking neural networks.
Approach
Spike-IMU employs a dynamic spiking neuron with an integer firing mechanism and a temporal feature fusion spike encoder to preserve signal fidelity, processed by a dynamic spiking LSTM to estimate velocity from raw IMU data.
Key results
- Novel dynamic spiking neuron with integer firing enhances representational bandwidth
- Temporal feature fusion spike encoder preserves short- and long-term motion dynamics
- Reduces positioning error by 20% compared to classical ANNs on the RoNIN dataset
- Achieves 70.3% lower energy consumption than RoNIN-TCN while maintaining superior accuracy
Why it matters
Enables high-accuracy, energy-efficient pedestrian navigation on resource-constrained edge devices like smartphones and wearables.
Abstract
Accurate pedestrian navigation on edge devices is a critical problem. While artificial neural networks (ANNs) have been shown to effectively solve this problem with acceptable accuracy, their energy consumption limits applications on low-power computation platforms. Spiking neural networks (SNNs) are promising alternatives, while their applicability in using noisy, high-frequency IMU data is hindered by two key issues: information loss during spike encoding and simplistic neuron dynamics that fail to capture complex motion. This paper introduces Spike-IMU, an SNN-based velocity estimation network designed to overcome these issues for the pedestrian navigation problem. In particular, a dynamic spiking neuron (DSN) is introduced based on the integer firing mechanism. In addition, a temporal feature fusion spike encoder (TFFSE) and a dynamic spiking long short-term memory network (DS- LSTM) are proposed to encode and process IMU data into spike sequences. Our experiments on the RoNIN dataset show that Spike-IMU surpasses classical ANNs, reducing positioning error by 20% while consuming 70.3% less energy. This work demonstrates a novel pipeline to design SNNs that achieves both superior accuracy and energy efficiency, pushing applications of IMU-based pedestrian navigation to real-world low-power edge devices.