# US 2021/0082270 A1 · Watching system, watching method, and recording medium

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0082270 A1
- Document ID: US20210082270A1
- Status: Published application
- Kind: A1
- Application number: 16/988,099
- Filed: 2020-08-07
- Published: 2021-03-18
- Earliest claimed: 2019-09-13
- Assignees: Hitachi, Ltd.
- Applicants: Hitachi, Ltd.
- Inventors: Yuto KOMATSU; Hirokazu ATSUMORI; Tsukasa FUNANE; Akiko OBATA; Akihiko KANDORI
- Web page: https://patentgrep.com/patent/US20210082270A1

## CPC

- G08B 21/043: {detecting an emergency event, e.g. a fall}
- A61B 5/1114: {Tracking parts of the body}
- A61B 5/1115: {Monitoring leaving of a patient support, e.g. a bed or a wheelchair}
- A61B 5/1117: {Fall detection}
- A61B 5/1128: {using image analysis (A61B5/1127 takes precedence)}
- A61B 5/6889: {Rooms}
- G08B 21/0476: {Cameras to detect unsafe condition, e.g. video cameras}
- G08B: SIGNALLING SYSTEMS, e.g. PERSONAL CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)

## Abstract

A watching system, which calculates a fall probability of an imaged person to be fallen using a computer which includes a processor and a memory, includes a trunk extraction unit that receives time-series skeleton information of the imaged person, calculates an angle between a head and a waist of the imaged person based on the skeleton information, extracts the angle as a trunk of the imaged person, calculates the center of the trunk as a trunk center based on trunk information set in advance, extracts a trunk to be reference as a reference trunk from a time-series trunk calculated from the time-series skeleton information, and holds a center of the reference trunk, and a fall probability calculation unit that calculates a distance between a trunk center of the reference trunk and a trunk center of the trunk as a deviation amount, and calculates the fall probability from the deviation amount.

## Claims

### Claim 1

A watching system for calculating a fall probability of an imaged person to be fallen using a computer which includes a processor and a memory, comprising:
a trunk extraction unit that receives time-series skeleton information of the imaged person, calculates an angle between a head and a waist of the imaged person based on the skeleton information, extracts the angle as a trunk the imaged person, calculates a center of the trunk as a trunk center based on trunk information set in advance, extracts a trunk to be reference as a reference trunk from a time-series trunk calculated from the time-series skeleton information, and holds a center of the reference trunk; and
a fall probability calculation unit that calculates a distance between a trunk center of the reference trunk and a trunk center of the trunk as a deviation amount, and calculates the fall probability from the deviation amount.

### Claim 2 (depends on 1)

The watching system according to claim 1, wherein
the trunk extraction unit calculates a time-series trunk center from the time-series trunk, and calculates an acceleration of the trunk center from the time-series trunk center, and
the fall probability calculation unit calculates the fall probability from the deviation amount of the trunk center and the acceleration.

### Claim 3 (depends on 2)

The watching system according to claim 2, wherein
the trunk extraction unit calculates an angle of legs from the time-series trunk, and
the fall probability calculation unit calculates the fall probability from the deviation amount of the trunk center, the acceleration, and the angle of the legs.

### Claim 4 (depends on 1)

The watching system according to claim 1, further comprising a warning generation unit that determines a sign of a fall when the fall probability is equal to or larger than a preset threshold.

### Claim 5 (depends on 4)

The watching system according to claim 4, wherein the warning generation unit outputs a predetermined warning when the sign of the fall is determined.

### Claim 6 (depends on 4)

The watching system according to claim 4, wherein the warning generation unit outputs a trunk correction image for correcting the trunk of the imaged person when the sign of the fall is determined.

### Claim 7 (depends on 4)

The watching system according to claim 4, wherein
the trunk extraction unit calculates a moving direction of the trunk center from the time-series trunk center as a falling direction, and
the warning generation unit is connected to a plurality of warning devices disposed around the imaged person, and generates a predetermined warning from the warning devices in a direction opposite to the falling direction when the sign of the fall is determined.

### Claim 8

A watching method for calculating a fall probability of an imaged person to be fallen using a computer which includes a processor and a memory, comprising:
a trunk extraction step in which the computer receives time-series skeleton information of the imaged person, calculates an angle between a head and a waist of the imaged person based on the skeleton information, extracts the angle as a trunk of the imaged person, calculates the trunk center as a trunk center based on trunk information set in advance, extracts a trunk to be reference as a reference trunk from a time-series trunk calculated from the time-series skeleton information, and holds a center of the reference trunk; and
a fall probability calculation step in which the computer calculates a distance between a trunk center of the reference trunk and a trunk center of the trunk as a deviation amount, and calculates the fall probability from the deviation amount.

### Claim 9 (depends on 8)

The watching method according to claim 8, wherein
in the trunk extraction step, a time-series trunk center is calculated from the time-series trunk, and an acceleration of the trunk center is calculated from the time-series trunk center, and
in the fall probability calculation step, the fall probability is calculated from the deviation amount of the trunk center and the acceleration.

### Claim 10 (depends on 9)

The watching method according to claim 9, wherein
in the trunk extraction step, the angle of the legs is calculated from the time-series trunk, and
in the fall probability calculation step, the fall probability is calculated from the deviation amount of the trunk center, the acceleration, and the angle of the leg.

### Claim 11 (depends on 8)

The watching method according to claim 8, further comprising a warning generation step for determining, by the computer, a sign of a fall when the fall probability is equal to or larger than a preset threshold.

### Claim 12 (depends on 11)

The watching method according to claim 11, wherein, in the warning generation step, a predetermined warning is output when the sign of the fall is determined.

### Claim 13 (depends on 11)

The watching method according to claim 11, wherein, in the warning generation step, a trunk correction image for correcting the trunk of the imaged person is output when the sign of the fall is determined.

### Claim 14 (depends on 11)

The watching method according to claim 11, wherein
in the trunk extraction step, a moving direction of the trunk center is calculated from the time-series trunk center as a falling direction, and
in the warning generation step, a predetermined warning is generated from a plurality of warning devices disposed around the imaged person in a direction opposite to the falling direction when the sign of the fall is determined.

### Claim 15

A non-transitory computer-readable recording medium having stored therein a program for causing a computer which includes a processor and a memory to execute a process comprising:
a trunk extraction step in which time-series skeleton information of the imaged person is received, an angle between a head and a waist of the imaged person is calculated based on the skeleton information, the angle is extracted as a trunk of the imaged person, the trunk center is calculated as a trunk center based on trunk information set in advance, a trunk to be reference is extracted as a reference trunk from a time-series trunk calculated from the time-series skeleton information, and a center of the reference trunk is held; and
a fall probability calculation step in which a distance between a trunk center of the reference trunk and a trunk center of the trunk is calculated as a deviation amount, and the fall probability is calculated from the deviation amount.
