Granted Patent B2

US 12,474,365 · App. 17/349,883 · Granted Nov 18, 2025

Inventors: Aditya Sarathy (Santa Clara, CA); Umamahesh Srinivas (Milpitas, CA); Bharath Narasimha Rao (San Mateo, CA); Alexander Singh Alvarado (San Jose, CA); Xiaoyuan Tu (Sunnyvale, CA); Jonathan Michel Beard (San Jose, CA)

Assignee: Apple Inc.

G01P15/02G01P13/04G06N3/08G06N7/01

Abstract

Claims (36)

1 . A method comprising:

training a classifier to detect human posture transition;

obtaining, with at least one processor, motion data from a headset reference frame associated with a headset worn by a user, the motion data including measurements of the user's inertial vertical acceleration and rotation about the user's torso in the headset reference frame;

transforming, with the at least one processor, the user's inertial vertical acceleration from the headset reference frame to a head/face reference frame;

determining, with the at least one processor, a dominant axis of the user's inertial vertical acceleration in the head/face reference frame;

determining, with the at least one processor, the user's face-forward inertial acceleration to be the dominant axis of the user's inertial vertical acceleration in the head/face reference frame;

determining, with the at least one processor, if the user's face-forward inertial acceleration meets or exceeds a minimum threshold;

in accordance with the user's face-forward inertial acceleration meeting or exceeding the minimum threshold, obtaining, with the at least one processor, one or more windows of motion data including the user's inertial vertical acceleration and the rotation about the user's torso;

detecting, with the trained classifier, a sit-to-stand or stand-to-sit posture transition of the user based on the one or more windows of motion data; and

determining, with the at least one processor, an activity of the user based on the user's posture transition.

2 . The method of claim 1 , wherein one or more phases of the user's posture transition include leaning, momentum and extension phases, and time boundaries of the extension phase are determined based on points of zero vertical inertial velocity derived from the user's inertial vertical acceleration, where the points of zero inertial vertical velocity correspond to maximum and minimum angles of the rotation about the user's torso.

3 . The method of claim 2 , wherein the user's posture transition is classified based on the one or more windows of motion data and additional motion data captured before and after each of the one or more windows of the motion data.

4 . The method of claim 1 , wherein the dominant axis is determined using principal component analysis.

5 . A system comprising:

one or more processors;

memory storing instructions that when executed by the one or more processors, cause the one or more processors to perform operations comprising:

training a classifier to detect human posture transitions;

obtaining motion data from a headset reference frame associated with a headset worn by a user, the motion data including measurements of the user's inertial vertical acceleration and rotation about the user's torso in the headset reference frame;

transforming the user's inertial vertical acceleration from the headset reference frame to a head/face reference frame;

determining a dominant axis of the user's inertial vertical acceleration in the head/face reference frame;

determining the user's face-forward inertial acceleration to be the dominant axis of the user's inertial vertical acceleration in the head/face reference frame;

determining if the user's face-forward inertial acceleration meets or exceeds a minimum threshold;

in accordance with the user's face-forward inertial acceleration meeting or exceeding the minimum threshold, obtaining one or more windows of motion data including the user's inertial vertical acceleration and the rotation about the user's torso; and

detecting, with the trained classifier, a sit-to-stand or stand-to-sit posture transition of the user based on the one or more windows of motion data; and

determining an activity of the user based on the user's posture transition.

6 . The system of claim 5 , wherein one or more phases of the user's posture transition include leaning, momentum and extension phases, and time boundaries of the extension phase are determined based on points of zero vertical inertial velocity derived from the user's inertial vertical acceleration, where the points of zero inertial vertical velocity correspond to maximum and minimum angles of the rotation about the user's torso.

7 . The system of claim 6 , wherein the user's posture transition is classified based on the one or more windows of motion data and additional motion data captured before and after each of the one or more windows of the motion data.

8 . The system of claim 5 , wherein the dominant axis is determined using principal component analysis.

9 . The method of claim 1 , further comprising:

determining, by inertial sensors of the headset and a companion device coupled to the headset, a gravity vector during a quiescence condition;

estimating a head/face to headset rotation transform based on the gravity vector; and

transforming, with the head/face to headset rotation transform, the user's inertial vertical acceleration from the headset reference frame to the head/face reference frame.

10 . The system of claim 5 , where the operations further comprise:

determining, by inertial sensors of the headset and a companion device coupled to the headset, a gravity vector during a quiescence condition;

estimating a head/face to headset rotation transform using based on the gravity vector; and

transforming, with the head/face to headset rotation transform, the user's inertial vertical acceleration from the headset reference frame to the head/face reference frame.

Assignments (1)

Continuity (2)

References Cited (52)