Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study introduces a wristband front end for continuous gesture sensing that integrates a MEMS IMU with a 60 GHz FMCW radar, achieving efficient performance within a coin-cell power budget. By employing a shared on-chip residual generator for classifier wake-up gating, routing, and EKF measurement reweighting, the design demonstrates significant energy savings and improved accuracy in gesture detection and pose tracking. The system achieves a detection probability of 0.72/0.80 at a 1% false-alarm rate while reducing classifier invocation energy by 47% and lowering pose tracking error by 4.6 times compared to traditional methods.
A wristband front end that cuts classifier energy use by nearly half while significantly enhancing gesture detection accuracy could revolutionize wearable gesture interfaces.
Continuous wrist-worn hand sensing for gesture interfaces and motor symptom monitoring needs an always-on front end that fits inside a coin-cell power budget while pairing a micro-electro-mechanical-systems (MEMS) inertial measurement unit (IMU) with a 60 GHz frequency-modulated continuous-wave (FMCW) radar to stay robust under occlusion and on-body drift. We present a design study of such a wristband front end in which classifier wake-up gating, mmWave versus IMU routing, and innovation-based EKF measurement reweighting share a single on-chip residual generator. The shared generator occupies 14.4 KB of program memory and 278 B of state and runs at 110K multiply-accumulates (MACs) per frame on an Ambiq Apollo4 Blue Plus class edge microcontroller unit (MCU). Across four public sensor data corpora (IPN Hand, SHREC 2021, MiliPoint 60 GHz FMCW radar, EAT-Radar) the front end reaches detection probability $P_D = 0.72/0.80$ at a 1% false-alarm rate, sustains a 47% classifier invocation energy reduction at 90% gesture detection recall, and lowers pose tracking root-mean-square error by $4.6\times$ under measurement bias drift relative to an adaptive Kalman with $R$-inflation baseline. Measured silicon power and on-body capture are deferred to follow-on hardware; the contribution here is a design study.