# US 11,298,050 B2 · Posture estimation device, behavior estimation device, storage medium storing posture estimation program, and posture estimation method > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,298,050 B2 - Document ID: US11298050B2 - Status: Granted - Kind: B2 - Application number: 16/750,471 - Filed: 2020-01-23 - Published: 2020-08-06 - Granted: 2022-04-12 - Earliest claimed: 2019-01-31 - Assignees: Konica Minolta, Inc. - Applicants: Konica Minolta, Inc. - Inventors: Naoki Ikeda - Web page: https://patentgrep.com/patent/US11298050B2 ## CPC - A61B 5/1113: {Local tracking of patients, e.g. in a hospital or private home} - A61B 5/1114: {Tracking parts of the body} - A61B 5/1116: {Determining posture transitions} - A61B 5/1117: {Fall detection} - A61B 5/1128: {using image analysis (A61B5/1127 takes precedence)} - G06T 7/11: Region-based segmentation - G06T 7/75: {involving models} - G06V 10/764: using classification, e.g. of video objects - G06V 40/103: {Static body considered as a whole, e.g. static pedestrian or occupant recognition} - G06V 40/23: {Recognition of whole body movements, e.g. for sport training} - G06T 2207/20084: Artificial neural networks [ANN] - G06T 2207/30196: Human being; Person - G06T 2207/30232: Surveillance - A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D) - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract A posture estimation device includes: a processor that: obtains detected information indicating a feature of a subject, wherein the feature is detected based on an image that is captured by an imager from a position where the subject is viewed from above, calculates a feature amount based on the obtained detected information, updates, based on a geometric relationship between the imager and the subject, a model parameter for estimating a posture of the subject by machine learning using the calculated feature amount in a time series, and estimates the posture of the subject using the updated model parameter. ## Claims ### Claim 1 A posture estimation device comprising: a processor that: obtains detected information indicating a feature of a subject, wherein the feature is detected based on an image that is captured by an imager from a position where the subject is viewed from above; calculates, based on the obtained detected information, a center-of-gravity distance between an upper body and a lower body of the subject; updates, based on a geometric relationship between the imager and the subject, a model parameter for estimating a posture of the subject by machine learning using the calculated center-of-gravity distance in a time series, wherein the geometric relationship is at least one of a position, orientation, and a posture of the subject with respect to the imager; estimates the posture of the subject using the updated model parameter, and detects, from the estimated posture, a change in a state or situation of the subject. ### Claim 2 (depends on 1) The posture estimation device according to claim 1, wherein the processor further: estimates the posture of the subject using a hidden Markov model, and updates the model parameter for calculating an output probability of the hidden Markov model based on the geometric relationship between the imager and the subject. ### Claim 3 (depends on 1) The posture estimation device according to claim 1, wherein the processor further calculates an index corresponding to the geometric relationship between the imager and the subject, and the processor further updates the model parameter for estimating the posture of the subject based on the calculated index. ### Claim 4 (depends on 1) The posture estimation device according to claim 1, wherein the detected information includes a human rectangle, a head rectangle, and a joint point detected from the image. ### Claim 5 (depends on 1) A behavior estimation device comprising: the posture estimation device according to claim 1, wherein the processor estimates behavior of the subject based on the estimated posture and detects the change from the estimated behavior. ### Claim 6 A non-transitory computer-readable storage medium storing a posture estimation program that causes a computer to: obtain detected information indicating a feature of a subject, wherein the feature is detected based on an image that is captured by an imager from a position where the subject is viewed from above; calculate, based on the obtained detected information, a center-of-gravity distance between an upper body and a lower body of the subject; update, based on a geometric relationship between the imager and the subject, a model parameter for estimating a posture of the subject by machine learning using the calculated center-of-gravity distance in a time series, wherein the geometric relationship is at least one of a position, orientation, and a posture of the subject with respect to the imager; estimate the posture of the subject using the updated model parameter, and detect, from the estimated posture, a change in a state or situation of the subject. ### Claim 7 (depends on 6) The non-transitory recording medium according to claim 6, wherein the program further causes the computer to: estimate the posture of the subject using a hidden Markov model, and update the model parameter for calculating an output probability of the hidden Markov model based on the geometric relationship between the imager and the subject. ### Claim 8 (depends on 6) The non-transitory recording medium according to claim 6, wherein the program further causes the computer to: calculate an index corresponding to the geometric relationship between the imager and the subject, and update the model parameter for estimating the posture of the subject based on the calculated index. ### Claim 9 (depends on 6) The non-transitory recording medium according to claim 6, wherein the detected information includes a human rectangle, a head rectangle, and a joint point detected from the image. ### Claim 10 A posture estimation method that causes a computer to execute a posture estimation program, the method comprising: obtaining detected information indicating a feature of a subject, wherein the feature is detected based on an image that is captured by an imager from a position where the subject is viewed from above; calculating, based on the obtained detected information, a center-of-gravity distance between an upper body and a lower body of the subject; updating, based on a geometric relationship between the imager and the subject, a model parameter for estimating a posture of the subject by machine learning using the calculated center-of-gravity distance in a time series, wherein the geometric relationship is at least one of a position, orientation, and a posture of the subject with respect to the imager; estimating the posture of the subject using the updated model parameter, and detecting, from the estimated posture, a change in a state or situation of the subject. ### Claim 11 (depends on 10) The posture estimation method according to claim 10, further comprising: estimating the posture of the subject using a hidden Markov model, and updating the model parameter for calculating an output probability of the hidden Markov model based on the geometric relationship between the imager and the subject. ### Claim 12 (depends on 10) The posture estimation method according to claim 10, wherein the method further comprising: calculating an index corresponding to the geometric relationship between the imager and the subject, and updating the model parameter for estimating the posture of the subject based on the calculated index. ### Claim 13 (depends on 10) The posture estimation method according to claim 10, wherein the detected information includes a human rectangle, a head rectangle, and a joint point detected from the image. ### Claim 14 (depends on 1) The posture estimation device according to claim 1, wherein the processor further: transmits, to a mobile terminal, event notification upon detecting the change. ### Claim 15 (depends on 1) The posture estimation device according to claim 1, wherein the processor further: detects the change as an event that requires event notification to a user, wherein the event includes at least one of getting up, getting out of bed, tumbling, falling, and abnormal micromotion. ### Claim 16 (depends on 6) The non-transitory recording medium according to claim 6, wherein the program further causes the computer to: detect the change as an event that requires event notification to a user, wherein the event includes at least one of getting up, getting out of bed, tumbling, falling, and abnormal micromotion. ### Claim 17 (depends on 10) The posture estimation method according to claim 10, wherein the detecting includes: detecting the change as an event that requires event notification to a user, wherein the event includes at least one of getting up, getting out of bed, tumbling, falling, and abnormal micromotion.