Using health monitor data to detect macro and micro habits with a behavioral model

Inventors

Hu, Julia • Zira, Jeff

Assignees

Lark Technologies Inc

Abstract

A method for communicating activity-related notifications to a user includes: receiving a record of activity events of a particular activity type performed by the user over a period of time; selecting a first time-based filter from a set of time-based filters; identifying a cluster of activity events in the record of activity events filtered according to the first time-based filter; identifying an early bound and a late bound on start times of activity events of the particular activity type from the cluster; communicating a notification of a first type to the user at a first time within a threshold time of the early bound on a day fulfilling the first time-based filter; and communicating a notification of a second type to the user at a second time within a threshold time of the late bound on a day fulfilling the first time-based filter.

Core Innovation

A remote computer system communicates activity-related notifications to a user by receiving a record of activity events of a particular activity type performed by the user over a period of time. The record specifies a start time and a duration of each activity event, and receiving the record comprises collecting location data sampled at a location sensor of a mobile computing device, where the location data comprises a location coincident performance of each activity event. The method selects a first time-based filter from a set of time-based filters and identifies a cluster of activity events in the record filtered according to the first time-based filter.

The cluster is identified by obtaining a set of computer-implemented rules defining activity event subgrouping as a function of activity event start time, activity event duration, and coincident location, where the set of computer-implemented rules is operable to improve the remote computing system in relation to behavioral pattern prediction. A subgroup of activity events is selected according to the computer-implemented rules based on composite start times and durations falling within a core composite distance threshold of a composite start time and a duration of at least one other activity event within the subgroup, and based on locations within a threshold distance of a location corresponding to at least one other activity event within the subgroup. The method predicts a representation of a behavioral pattern of the user in the subgroup based on identification of the cluster.

From the cluster, the method identifies an early bound and a late bound on start times of activity events of the particular activity type. The method communicates a notification of a first type at a first time succeeding the period of time, where the first time is within a threshold time of the early bound on a day fulfilling the first time-based filter. The method also communicates a notification of a second type at a second time succeeding the period of time, where the second time is within a threshold time of the late bound on a day fulfilling the first time-based filter, with the second notification occurring when activity is absent between the early bound and the late bound on the day fulfilling the first time-based filter.

In a further implementation, the method extrapolates from the cluster a quantitative datum corresponding to the particular activity type and the first time-based filter, and at a first time within a threshold time of the early bound presents a form of the quantitative datum through the display of the mobile computing device. At a second time within a threshold time of the late bound, in response to an absence of a detected activity of the particular activity type between the early bound and the late bound on a day fulfilling the first time-based filter, the method presents to the user a prompt to provide feedback for the absence of a detected activity of the particular activity type.

Claims Coverage

The provided independent claims are clm-00001 and clm-00017. Each includes the same overall inventive architecture: receiving time-bounded activity-event records with coincident location, selecting a time-based filter, clustering via computer-implemented rules using time/duration/location relationships, predicting a behavioral pattern representation, deriving early and late bounds, and triggering different user communications around those bounds; claim clm-00017 additionally extrapolates a quantitative datum and ties the second communication to prompting user feedback for an absence within the bound window.

Together, the independent claims cover selecting time-based filters for activity-event records with coincident location, clustering and subgrouping activity events using computer-implemented rules based on composite start time, duration, and location thresholds, predicting a behavioral pattern representation and deriving early and late bounds, and communicating different notification types around those bounds. Claim clm-00017 further adds extrapolating a quantitative datum for early-bound presentation and prompting user feedback when activity is absent between the early and late bounds.

Stated Advantages

Documented Applications