Wrist-Worn IMU Walking Speed Estimation via Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for measuring walking speed, such as clinical tests and wearable sensors, face limitations in accuracy, applicability, and convenience, especially for longitudinal monitoring in real-world settings, due to issues like velocity drift in inertial sensors and the need for specific mounting positions.
Innovation Solution
A system and method using a wrist-worn inertial measurement unit (IMU) with sensor fusion and principal component analysis to process acceleration and rate of turn signals, generating a principal component acceleration signal that improves walking speed estimation through regression-based models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If integration-based approaches are used to track velocity from accelerometer data, then walking speed measurement is enabled in ambulatory fashion, but velocity drift occurs over time due to time-varying bias in MEMS-based inertial sensors
Solution Approach 1:
The system uses feedback by detecting periodic foot stance phases during walking to reset the velocity to zero through a process called zero velocity update (ZUPT). This feedback mechanism continuously corrects the velocity drift by utilizing the known physical constraint that velocity is zero during foot stance phases, thereby maintaining measurement precision over time while preserving ambulatory operation capability
Solution Approach 2:
The system changes the operational parameters by switching between integration-based velocity tracking and zero-velocity resetting based on detected foot stance phases. This parameter change approach allows the system to maintain accurate velocity measurements by periodically resetting the integration accumulator to zero during stance phases, thereby eliminating cumulative drift errors while preserving continuous ambulatory measurement capability
2Measurement precision
If zero velocity update (ZUPT) is used to mitigate drift by resetting velocity to zero, then measurement precision is improved, but the wearable sensor must be mounted on the leg (ideally on the foot) which is inconvenient for longitudinal monitoring
Solution Approach 1:
The system achieves universality by developing a waist-mounted IMU configuration that can perform both gait event detection and walking speed estimation functions. The waist location serves as a universal mounting position that enables convenient longitudinal monitoring while still allowing the system to detect foot stance phases through hip motion patterns and perform velocity resetting, thereby maintaining measurement precision without requiring inconvenient foot-mounted sensors
Solution Approach 2:
The system uses the hip as an intermediary between the waist-mounted sensor and the foot. By detecting hip motion patterns that correlate with foot stance phases, the system can indirectly detect gait events without direct foot contact. This intermediary approach allows the waist-mounted sensor to infer foot stance timing through hip kinematics, enabling velocity resetting at the appropriate moments while maintaining convenient waist-mounted operation
3Measurement precision
If camera-based systems are used for in-home gait speed measurement, then measurement capability is provided, but the systems are affected by lighting conditions and limited to confined areas
Solution Approach 1:
The system replaces the optical/camera-based measurement approach with a mechanical inertial sensing approach using MEMS accelerometers and gyroscopes. This substitution eliminates dependence on lighting conditions and camera field of view, allowing the system to function independently of environmental optical conditions. The inertial sensors provide autonomous measurement capability that is not affected by lighting variations or spatial confinement, thereby significantly improving environmental adaptability while maintaining gait speed measurement precision
Data Source
AI summary
Systems and methods are provided for estimating a walking speed of a subject. A method comprises mounting an inertial measurement unit (IMU) on a wrist of the subject, the IMU configured to generate acceleration and rate of turn signals; processing the acceleration and rate of turn signals from the IMU to generate a pitch angle and a roll angle; processing the pitch angle and the roll angle to generate a rotation matrix from a sensor frame of the IMU to a navigation frame of the subject; applying the rotation matrix to the acceleration signals and removing gravitational acceleration to generate an external acceleration signal; processing the external acceleration signal to determine a principal horizontal axis and to generate a principal component acceleration signal representing external acceleration along the principal horizontal axis; and processing the principal component acceleration signal using a regression-based method to determine an estimated walking speed of the subject.