Multimodal Wearable Sensors to Measure Gait and Voice.

BACKGROUND

Traditional measurement systems utilized in clinical trials are limited because they are episodic and thus cannot capture the day-to-day fluctuations and longitudinal changes that frequently affect patients across different therapeutic areas.

OBJECTIVES

The aim of this study was to collect and evaluate data from multiple devices, including wearable sensors, and compare them to standard lab-based instruments across multiple domains of daily tasks.

METHODS

Healthy volunteers aged 18-65 years were recruited for a 1-h study to collect and assess data from wearable sensors. They performed walking tasks on a gait mat while instrumented with a watch, phone, and sensor insoles as well as several speech tasks on multiple recording devices.

RESULTS

Step count and temporal gait metrics derived from a single lumbar accelerometer are highly precise; spatial gait metrics are consistently 20% shorter than gait mat measurements. The insole's algorithm only captures about 72% of steps but does have precision in measuring temporal gait metrics. Mobile device voice recordings provide similar results to traditional recorders for average signal pitch and sufficient signal-to-noise ratio for analysis when hand-held. Lossless compression techniques are advised for signal processing.

CONCLUSIONS

Gait metrics from a single lumbar accelerometer sensor are in reasonable concordance with standard measurements, with some variation between devices and across individual metrics. Finally, participants in this study were familiar with mobile devices and had high acceptance of potential future continuous wear for clinical trials.