Ground Truth:

Reconstructed using 15
IMU (XSens MVN)

Inference:

Using 3 Sparse IMU
of XSens MVN.

Inference:

Using 3 Standalone
Low-cost IMUs .

Overview

For upper extremity rehabilitation, quantitative measurements of a person’s capabilities during activities of daily living could provide useful information for therapists, including in telemedicine scenarios. Specifically, measurements of a person’s upper body kinematics could give information about which arm motions or movement features are in need of additional therapy, and their location within the home could give context to these motions. To that end, we present a new algorithm for identifying a person’s location in a region of interest based on a Bluetooth received signal strength (RSS) and present an experimental evaluation of this and a different Bluetooth RSS-based localization algorithm via fingerprinting. We further present algorithms for and experimental results of inferring the complete upper body kinematics based on three standalone inertial measurement unit (IMU) sensors mounted on the wrists and pelvis.

Overview of our strategy for in-home localization and kinematics monitoring. (a) In-home localization approach. Bluetooth transmitters are installed around the home, and the received signal strength is monitored at the patient. The localization provides context to the captured activities. (b) Minimal sensor set that is unobtrusive during daily life, including IMUs worn on each wrist and the waist, and a Bluetooth receiver worn at the waist. (c) Kinematic reconstruction of the torso derived from the worn IMUs.

Data Collection Process

Data collection using both XSens MVN suit (ground truth) and XSens DOT (on LFA,RFA,Pelvis,LSH) We created two distinct sets: {LFA, RFA, Pelvis} and {LFA, RFA, LSH}. Within each set, we conducted an analysis to assess the performance of upper-body kinematics reconstruction.

Sample Daily Activity Recorded. Stick figure represents reconstructed pose (Ground Truth) of upperbody using 15 IMU of XSens MVN.

Data Processing & Deep Learning Pipeline

Data collection using both XSens MVN suit (ground truth) and XSens DOT (on LFA,RFA,Pelvis,LSH) We created two distinct sets: {LFA, RFA, Pelvis} and {LFA, RFA, LSH}. Within each set, we conducted an analysis to assess the performance of upper-body kinematics reconstruction.

Quantitative Results

Distribution of the mean angular error for motion inference (for all four models) using the sparse segments of XSens MVN (from the new dataset).

Qualitative Results

Data collection using both XSens MVN suit (ground truth) and XSens DOT (on LFA,RFA,Pelvis,LSH) We created two distinct sets: {LFA, RFA, Pelvis} and {LFA, RFA, LSH}. Within each set, we conducted an analysis to assess the performance of upper-body kinematics reconstruction.

Remarks

On this page, I present a brief highlights of the upper body kinematics reconstruction using our proposed method, which was my primary contribution. Other aspects of the work, such as localization using RSSI, quantitative results are not presented here. For more details, please refer to the paper. This research was funded by the National Science Foundation (Grant # 2014499)

In collaboration with Carnegie Mellon University, we are currently developing new sensors that can simultaneously provide IMU and RSSI information. Our near future goal is to use these sensors to reconstruct the upper body kinematics and localize the person in the home.

BibTeX