Detecting falls with multiple wearable motion sensors

Inventors

Filatov, Denis • Zaveri, Hitten

Assignees

Alva Health Inc

Abstract

In an example system, multi-sensor motion data that reflects the motion of a user over a period of time is received as recorded by a plurality of motion sensors on wearable devices, specific changes are determined in the data by, firstly, quantifying it using a two-dimensional data transform; secondly, extracting an anomalous area; and thirdly, calculating a number of the anomaly's properties, and the results are input into a machine learning model to detect that the user fell. The machine learning model processes the anomaly's properties and evaluates the current state of the user's activity by classifying the properties against a state space previously calculated by analyzing historical activities of daily living. Depending on a two-dimensional transform implemented, the machine learning model detects when the user falls, as well as potentially allows predicting that the user will suffer a fall in advance of the actual event, aiming at solving the stroke prediction problem.

Core Innovation

The invention receives, at a plurality of wearable devices worn by a user, multi-sensor motion data based on the user's activities over a period of time, where each wearable device includes a motion sensor and includes a first wearable device worn on a first wrist and a second wearable device worn on a second wrist. The invention calculates one or more properties of the multi-sensor motion data by filtering time series of the multi-sensor motion data and calculating a measure of coherence of the filtered multi-sensor motion data. Using the obtained measure of coherence, the invention determines an area in which coherence's values exceed a threshold and calculates one or more properties of the area.

The invention determines that the user fell using a machine learning model executed on a processor of a cloud service and based on the one or more properties of the multi-sensor motion data, where the machine learning model is previously trained on historical multi-sensor motion data. The invention transmits the determination to a server, a caretaker of the user, and an emergency system. The invention further characterizes the data processing through calculating properties of coherence-exceeding areas derived from the multi-sensor motion data.

The invention is also implemented as a computer program product on a non-transitory computer readable medium comprising computer instructions to perform the receiving, filtering, coherence-measure calculation, area determination, property calculation, and cloud-based machine learning fall determination. It also includes a system implementation with a processor configured to perform these operations and a memory coupled to the processor to provide the processor with instructions.

Claims Coverage

The independent claims are method, computer program product, and system. Across these independent claims, the same core claim pattern is present: multi-wrist multi-sensor motion data, filtering, coherence measure, thresholded coherence-area, property calculation, cloud-executed machine learning for falling determination, and transmission to server, caretaker, and emergency system.

All independent claims cover determining a fall from multi-sensor motion data collected by first and second wrist-worn devices by computing a coherence measure, defining a thresholded coherence-exceeding area, calculating properties of that area, and using a cloud-executed machine learning model previously trained on historical multi-sensor motion data to decide the fall, followed by transmission of the determination to server, caretaker, and emergency system.

Stated Advantages

Documented Applications