Computer-implemented method for fall assessment and actuation implementing a trained machine learning model

A machine learning-based fall assessment method with continuous data processing and personalized protection mechanisms addresses practicality and privacy issues in existing fall detection, ensuring timely and effective responses to falls.

JP2024543802APending Publication Date: 2024-11-26AUTOLIV DEV AB
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
JP2024524666
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2021-10-29
Filing Date
2022-10-28
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Existing fall detection and prediction methods, particularly in external environments, face practicality issues and privacy concerns with vision-based solutions, and wearable device sensors lack effective customization for immediate and targeted protection during falls.

Method used

A computer-implemented method utilizing a trained machine learning model for fall assessment that includes continuous data acquisition, processing, and actuation, enabling personalized physical protection through airbag deployment and alarms based on fall characterization and context, using sensors like accelerometers and gyroscopes.

Benefits of technology

Provides rapid, customizable protection and early warning for falls, optimizing actuation based on fall type, severity, and impact time, enhancing user safety by minimizing injury risk.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024543802000001_ABST
    Figure 2024543802000001_ABST
Patent Text Reader

Abstract

The present disclosure relates to a computer-implemented method for fall assessment for a fall assessment environment, the computer-implemented method including implementing a trained machine learning model including a behavior plan and an actuation according to the behavior plan, the behavior plan being generated reflective of training of the machine learning model. The actuation according to the action plan includes protective measures and reflects a fall assessment for the user (1), the fall assessment including fall detection and fall characterization of the detected fall. The computer-implemented method includes continuously acquiring data and continuously processing the data, the continuously acquired data including data detected by sensors (2, 3, 6, 7) worn by the user (1).
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates to an improved computer-implemented method for fall assessment, a computer program comprising computer readable instructions for applying the computer-implemented method, a computer program comprising computer readable instructions for a trained machine learning model, a computer program comprising computer readable instructions for a "fall characterization machine learning unit", a computer readable medium comprising the computer program, and a control unit arrangement, device and system for user fall assessment. [Background technology]

[0002] Falls are one of the leading causes of human injuries and injury-related deaths. Fall detection and fall prediction are performed with machine learning approaches of increasing usefulness. Machine learning approaches are used today, for example, to distinguish between falling and non-falling movements. At a high level, these approaches may include approaches with vision-based configurations that can classify falls in real time, based on live feed video. Vision-based configurations may be less practical in some cases, for example in external environments, and may involve potential privacy issues. Furthermore, there are machine learning approaches that focus on more pervasive solutions based on wearable device sensors, which are non-intrusive and pervasive. In this field, see for example "A Cascade-Classifier Approach for Fall Detection, Putra et al, MOBIHEALTH 2015, October 14-16, London, Great Britain, January 2015, DOI:10.4108 / eai.14-10-2015.2261619," which proposes a cascade-classifier approach for this purpose. Other examples include M. Musci, D. D. Martini, N. Blago, T. Facchinetti, and M. Piastra, "Fall Detection using Recurrent Neural Networks," p. 7, as well as F. J. Gonzalez-Canete and E. Casilari, "A Feasibility Study of the Use of Smartwatches in Wearable Fall Detection Systems," Sensors, vol. 21, no. 6, p. 2254, Mar. 2021, doi:10.3390 / s21062254.

[0003] WO21050966 discloses a system and method for predicting and preventing impending falls of residents of an institution (e.g., a hospital, a care home, or a home) using sensors. The system 100 may also be used to help prevent falls for residents in general, and the microphone may be used to determine information regarding the type of fall, the degree of severity of the fall.

[0004] A fall may occur, for example, to a person walking in a domestic environment or outside the home. In this context, a fall also includes falls that occur while riding a bicycle, scooter, motorcycle, etc., due to, for example, skidding, slipping, loss of control of the vehicle, crashing with another vehicle or object, etc. Falls involving a moving vehicle are often more abrupt and require more immediate attention than falls that occur to a person walking, for example, by tripping or slipping. Summary of the Invention [Problem to be solved by the invention]

[0005] However, there is still room for improvement in this area, particularly with regard to the requirement for improved methods for fall assessment.

[0006] This is accomplished by a computer-implemented method for fall assessment for a fall assessment environment, the computer-implemented method including implementing a trained machine learning model including a behavior plan and an actuation according to the behavior plan, the behavior plan being generated reflective of training of the machine learning model.

[0007] The actuation according to the action plan includes protective measures and reflects a fall assessment for the user, the fall assessment including fall detection and fall characterization of the detected fall. The computer-implemented method includes continuously acquiring data and continuously processing the data, the continuously acquired data including data detected by a sensor worn by the user.

[0008] This means that a user's customized physical protection can be provided using an action plan generated depending on the context scenario, for example predicting the time before impact and which body area after the collision (head, elbow, hip, side) will have the maximum impact, and then triggering physical protection accordingly.

[0009] According to some aspects, the protective measures include physical measures including inflating an airbag.

[0010] This provides a rapidly deployable type of protection that can be installed in a suitable location.The actuation may for example include the mechanical actuation of a portable airbag after detection of a rollover incident.

[0011] According to some aspects, actuation according to the action plan then includes measures including at least one of preventative measures and alarm measures.

[0012] Warning measures are suitable when the time to impact is too short to take any protective measures, for example when a fall event is detected too late.

[0013] This means that actuation includes measures such as protective and / or alarm measures, where protective measures include physical measures such as inflating an airbag as soon as it is detected that a person is in a dangerous position vulnerable to a fall, such as in a pre-fall situation.

[0014] According to some aspects, the computer implemented method includes a continuous generation of information based on the acquired data and the processed data, the computer implemented method includes a continuous communication of the information. The computer implemented method includes a fall assessment for the user, the fall assessment including fall detection and fall characterization based on the continuous data and the continuous information.

[0015] According to some aspects, as described herein, a computer-implemented method for fall assessment for a fall assessment environment includes a trained machine learning model trained in fall assessment, a user, a sensor for detecting data from the user's fall, and a means for continuously acquiring data. The fall assessment environment further includes a means for processing the data, a means for generating information based on the acquired data and the processed data, a means for communicating the information, and a means for actuation according to an action plan and based on the sequential data and sequential information.

[0016] The trained machine learning model, which is trained in a fall assessment, includes, during its implementation in the computer-implemented method for fall assessment, continuously acquiring data and continuously processing the data; The continuous creation and communication of information; Fall assessment for a user, including fall detection and fall characterization; and It enables continuous actuation that follows an action plan and reflects fall assessment.

[0017] According to some aspects, a computer-implemented method according to the present disclosure is disclosed, as described herein, in which the trained machine learning model continuously enables estimation of the confidence in any prediction in the computer-implemented method during its implementation in the computer-implemented method for fall assessment. The computer-implemented method includes estimating the confidence in any prediction of the fall characterization and selecting an action plan and actuation according to the prediction if the confidence in the prediction is high. If the confidence in the prediction is not high, default actuation according to the action plan.

[0018] As an example, there may be a high degree of confidence that a fall has occurred or is about to occur, but there may still be a low degree of confidence in the characterization of the fall.

[0019] According to some aspects, a computer-implemented method according to the present disclosure is disclosed as described herein, the computer-implemented method including predicting a severity of a detected fall.

[0020] This can provide input as to which protective and preventative measures should be activated, and to what extent.

[0021] According to some aspects, a computer-implemented method according to the present disclosure as described herein is disclosed, wherein the trained machine learning model continuously enables estimation of the time to impact of a detected fall during its implementation in the computer-implemented method for fall assessment, the computer-implemented method including estimating the time to impact of a detected fall.

[0022] This means that factors such as, for example, the characterization of the fall, the severity of the fall, and the time to impact are calculated and taken into account for the action plan for actuation, which can provide input as to which protective and preventative measures should be activated, and to what extent.

[0023] The present disclosure also relates to computer programs, control units, devices, systems, and garments associated with the above advantages.

[0024] It is therefore understood that the computer-implemented methods, computer programs, and computer-readable media all described herein and according to the present disclosure may be realized in hardware, such as the control unit configurations and devices all described herein, and in the systems described herein. The hardware, such as the control unit configurations and devices all described herein, and the systems described herein are then configured to execute the computer-implemented methods and computer programs, thereby obtaining the same advantages and benefits as discussed herein for the computer-implemented methods. [Brief description of the drawings]

[0025] The present disclosure will now be described in more detail with reference to the accompanying drawings. [Figure 1] FIG. 1 illustrates generally a walking user wearing a sensor. [Diagram 2] FIG. 2 illustrates diagrammatically a user riding a motorbike and wearing a sensor. [Diagram 3] FIG. 3 illustrates a schematic diagram of aspects of a “customized safety strategy for fall incidents” according to the present disclosure, i.e., action plan and actuation by the action plan computer-implemented method for fall assessment. [Figure 4] FIG. 4 illustrates diagrammatically the control unit. [Diagram 5] FIG. 5 illustrates an exemplary computer program. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0026] Aspects of the present disclosure will now be described more fully below with reference to the accompanying drawings. However, the different configurations, devices, systems, computer programs, and methods disclosed herein may be embodied in many different forms and should not be construed as being limited to the aspects set forth herein. Like numbers in the drawings refer to like elements throughout.

[0027] The terms used herein are for the purpose of describing aspects of the present disclosure only and are not intended to be limiting of the present invention. As used herein, the singular forms "a," "an," and "the" may also be intended to include the plural forms unless the context clearly indicates otherwise.

[0028] 1, there is a user 1 wearing at least one sensor device 2, 3, here one wrist sensor device 2 and one waist sensor device 3. In this case, the user 1 is walking.

[0029] 2, there is a user 1 wearing at least one sensor device, here one wrist sensor device 6 and one vest sensor device 7, included in a protective garment such as a protective vest 10. In this case the user 1 is riding a motorbike 8.

[0030] The present disclosure is applicable to users moving on foot or on or in any type of vehicle, such as a bicycle, motorbike 8, car, etc.

[0031] The acquired data includes data acquired from sensors 2, 3, 6, 7 according to the present disclosure, and may include data acquired from other sources. The acquired data from sensors 2, 3, 6, 7 is typically collected, i.e. acquired, from sensors included, e.g. worn, in a wearable device, e.g. motion sensors, e.g. 3d sensors such as accelerometers for measuring acceleration in three dimensions, as well as gyroscopes for measuring rotational speed in three dimensions from a person 1, i.e. a user, according to the present disclosure, wearing such a wearable device. The wearable devices 2, 3, 6, 7 may also suitably be edge devices. The acquired data may be measurements from other sensors, such as magnetometers. A magnetometer is a device that measures magnetic fields. Furthermore, the acquired data may also include measurements of the strength or relative change of a magnetic field at a particular location.

[0032] The acquired data may include data acquired from sources other than sensors. The acquired data may refer to current and past historical data. This may include the user's geographic location information from a GPS system as well as weather information. Further information sources may include static or semi-static information such as the user's physical profile, such as age, height, BMI, any physical disabilities, and any relevant medical information, such as vulnerability to falls, among other conditions, that is withdrawn from the historical data, etc. Such personal information of the user may be used in the computer-implemented method for fall assessment, as described herein, to fine-tune the fall severity calculation for that particular user.

[0033] According to some embodiments, the user 1 wears means 11 for providing and communicating the location of the user 1. Such means 11 may comprise any type of suitable positioning system, such as for example GPS or GNSS (Global Navigation Satellite System), and any type of wireless communication system. The means may be included in a vest 11 or any suitable type of garment or wearable device.

[0034] The present disclosure relates to a computer-implemented method for fall assessment for a fall assessment environment, the computer-implemented method including implementing a trained machine learning model including an action plan and an actuation according to the action plan, the action plan being generated reflecting training of the machine learning model.

[0035] Actuation according to the action plan includes protective measures and reflects a fall assessment for the user 1, the fall assessment including fall detection and fall characterization of the detected fall. The computer-implemented method includes continuously acquiring data and continuously processing the data, the continuously acquired data including data detected by sensors 2, 3, 6, 7 worn by the user 1.

[0036] This means that customized physical protection of user 1 can be provided using an action plan generated depending on the context scenario, for example predicting the time before impact and which body area after the collision (head, elbow, hips, side) will have the maximum impact, and then triggering physical protection accordingly.

[0037] According to some aspects, the protective measures include physical measures including inflating an airbag.

[0038] This provides a rapidly deployable type of protection that can be attached in a suitable location. Moreover, this means that the actuation can include, for example, a mechanical actuation of the portable airbag 4, 5, 9 after detection of a fall incident. The safety capacity of the portable airbag 4, 5, 9 and its properties such as volume, inflation capacity and inflation rate are based on an action plan and guidance according to the action plan. The guidance includes actuation and preventive measures and correlates and / or reflects a fall assessment for the user, which includes fall detection and calculation of the FRP for that particular user.

[0039] According to some aspects, actuation according to the action plan then includes measures including at least one of preventative measures and alarm measures.

[0040] Warning measures are suitable when the time to impact is too short to take any protective measures, for example when a fall event is detected too late.

[0041] According to some aspects, the actuation includes measures such as protective measures and / or warning measures, where warning measures include alerting a user of the computer-implemented method for fall assessment as described herein when the user is in a dangerous situation, e.g., before a pre-fall situation, and when the user may be at risk of being injured or injured. Thus, the user may then be able to quickly take some corrective measures and / or any walking posture corrections, for example, to avoid potential injury. The warning measures may be in the nature of a visual notice, such as on a small screen or using a light bulb, an audible warning, and / or a haptic feedback pattern, to alert the user of an upcoming potentially dangerous situation.

[0042] In this context, physical measures are part of protective measures that in turn are concerned with protecting the person in the event of a fall. Preventive measures are measures to prevent falls, such as issuing an alarm when there is a high risk of a fall occurring. In the case of motorcyclists, this could be a notification issued dependent on slippery / icy road conditions detected.

[0043] According to some embodiments, the preventative measures include a warning signal and the protective measures include the inflation of one or more protective airbags 4, 5, 9 as illustrated in Figures 1 and 2. As shown in Figure 2, for a person or user 1 riding a motorbike 8 or similar, the sensors 6, 7, control unit 700, and / or protective airbag 9 may all be included in protective clothing such as a protective vest 10.

[0044] Moreover, preventative measures refer to measures to avoid a fall incident in the first place, before it occurs, including, but not limited to, actual warnings to the user, such as visual, audible, and / or tactile warnings, when a dangerous position is detected, such as a position close to a fall (but not yet at the point of no return).

[0045] Additionally, preventative measures may also include other physical preventative measures, such as increasing protective clothing and / or garments for the user, and / or may further include increasing monitoring of the user. Additionally, these preventative measures may also be based on fall assessment, which may include fall detection and calculation of an FRP for a particular user, which may look at calculating the FRP in conjunction with GPS geographic positioning, as described herein, for example, calculating the FRP in conjunction with the geographic positioning of a user currently utilizing GPS.

[0046] A computer-implemented method for fall assessment, as described herein, includes actuation according to an action plan, the action plan having been generated reflecting training of a machine learning model, the actuation reflecting a fall assessment for a user, the fall assessment including fall detection and fall characterization.

[0047] The action plans are generated reflecting the training of machine learning models, and the generation of the action plans is based on the output of different trained machine learning models that predict characteristics such as the type of fall that is likely to occur (if a fall is predicted to occur) and the severity of the fall based on factors such as the type of fall and acceleration / velocity of impact with the ground / impact.

[0048] Fall detection is the first step to detect a fall and trigger other mechanisms to generate action plans and preventative and protective measures. Fall detection is performed in real-time using a trained machine learning model on collected data, i.e. data obtained from a user, e.g. a trainer equipped with a 3D sensor. The trained fall detection machine learning model generates a signal with a certain lead time if a fall is approaching, based on the current and past state data of the user equipped and using the 3D sensor.

[0049] Fall characterization refers to discriminatory features such as fall type characterization, such as, for example, frontal fall, sideways fall, backward fall from a slip, etc., among others. Methods for this may be based on data-driven approaches such as machine learning, or may be based on analytical methods based on domain knowledge.

[0050] Another important feature of fall characterization is a method for classifying the severity of a fall that correlates with the probability of serious injury from the fall. This can be based on data collected in real time from sensors, but also on historical data of the user based on aspects such as medical history, disease status, and age.

[0051] According to some aspects, the computer implemented method includes a continuous generation of information based on the acquired data and the processed data, the computer implemented method includes a continuous communication of the information. The computer implemented method includes a fall assessment for the user, the fall assessment including fall detection and fall characterization based on the continuous data and the continuous information.

[0052] The computer-implemented method includes a fall assessment for a user, the fall assessment including fall detection and fall characterization based on continuous data and continuous information, as described herein.

[0053] The computer-implemented method includes fall detection based on continuous data and continuous information as described herein, where a fall detection machine learning model is trained on real-time data collected from a 3D sensor.

[0054] Additionally, a fall characterization computer program is used that includes instructions for a "fall characterization machine learning unit."

[0055] The computer-implemented method further includes fall characterization based on the sequential data and sequential information as described herein. The fall characterization is triggered only if a fall is first detected by the fall detection machine learning model. The intent of the computer-implemented method including fall characterization is to detect and / or predict the type of fall and / or the nature of the fall.

[0056] Further, according to some embodiments, a fall severity computer program including instructions for a "fall severity prediction machine learning module" is also used in connection with fall type prediction. The fall severity computer program detects the severity of a fall or the probability of significant injury from a particular fall, e.g., a particular fall type. This fall severity computer program can take as input, i.e., obtain, real-time data streaming from the 3D sensor, as well as historical data from the user and its user profile and other information to estimate the probability of significant injury from a fall impact for a particular user. If both of these characteristics can be predicted with high confidence and / or probability, this allows for the generation of a customized action plan according to the needs of the particular user in that fall scenario when a fall is first detected.

[0057] According to some aspects, the fall assessment environment comprises a trained machine learning model trained in fall assessment, a user 1, sensors 2, 3, 6, 7 for detecting data from the user's fall, means for continuously acquiring data, and means for processing data 700. The fall assessment environment further comprises means for generating information based on the acquired and processed data, means for communicating the information, and means for actuation according to an action plan and based on the successive data and the successive information. The trained machine learning model, during its implementation in the computer-implemented method for fall assessment, -Continuous acquisition of data and continuous processing of the data; -Continuous creation and communication of information, - fall assessment for user 1, including fall detection and fall characterization; and -Enables continuous actuation according to action plans and reflective of fall assessment.

[0058] The computer-implemented method may further include, as described herein, the “fall characterization machine learning unit” comprising a “fall type prediction machine learning module”, which during said implementation of the trained machine learning model, when a fall is detected, performs: Predicting the fall type of a detected fall, and A computer-implemented method for fall assessment includes fall characterization of a detected fall, the fall characterization comprising: predicting a fall type; This allows for inclusion of action plans and actuation according to the predicted fall type.

[0059] Further, prediction of the type of fall, along with the associated severity with high confidence, following fall detection enables the proposed computer-implemented method for fall assessment, as described herein, to create an optimized actuation plan customized for a particular user under that particular fall incident dynamics, the optimized actuation plan including, for example, actuation such as actuation of a portable airbag in a particular direction based on the type of fall as well as other factors such as depending on the severity of the fall, and adjustment of actuation parameters such as inflation speed and / or volume or trigger of multiple portable airbags depending on the type of fall and its severity for a particular user.

[0060] This is driven by the action plan. Time to impact provides further feedback to assess whether there is sufficient time available to perform certain actuations before impact occurs and damage is potentially sustained. The primary objective of the computer-implemented method for fall assessment is to maximize the protection of the user from serious injury in the event of a fall incident, depending on the user profile, medical conditions, fall dynamics, and environmental conditions, as described herein.

[0061] The computer-implemented method further includes fall characterization, where the fall characterization includes predicting a fall type of the detected fall.

[0062] According to some aspects, the trained machine learning model continuously enables, during its implementation in the computer-implemented method for fall assessment, an estimation of the confidence in any prediction of the fall characterization in the computer-implemented method, the computer-implemented method including estimating the confidence in any prediction, and performing an action plan and actuation selection according to the prediction if the confidence in the prediction is high, and performing a default actuation according to the action plan if the confidence in the prediction is not high.

[0063] The characterization of the fall may include, for example, the type and / or severity of the fall. As an example, there may be a high degree of confidence that a fall has occurred or is about to occur, but still a low degree of confidence in the characterization of the fall.

[0064] According to some aspects, the computer-implemented method includes predicting a severity of the detected fall.

[0065] This can provide input as to which protective and preventative measures should be activated, and to what extent.

[0066] According to some aspects, the trained machine learning model continuously enables estimation of a time to impact of a detected fall during its implementation in a computer-implemented method for fall assessment, the computer-implemented method further including estimating a time to impact of a detected fall.

[0067] This means that factors such as, for example, the characterization of the fall, the severity of the fall, and the time to impact are calculated and taken into account for the action plan for actuation, which can provide input as to which protective and preventative measures should be activated, and to what extent.

[0068] In a further aspect of the disclosure, as described herein, characterizing the fall includes predicting a fall type of the detected fall, and characterizing the fall further includes estimating a confidence in any prediction and / or predicting a severity of the detected fall.

[0069] According to some aspects, the fall characterization includes prediction of a fall type, and actuation is performed according to the action plan and the predicted fall type.

[0070] For example, a proposed solution according to a computer-implemented method for fall assessment as described herein is depicted in FIG. 3. FIG. 3 relates to an aspect of a "customized safety measure for fall incidents" according to the present disclosure, i.e., action plan and actuation according to the action plan computer-implemented method for fall assessment. In FIG. 3, it can be seen that data is streamed in real-time from sensors 2, 3, 6, 7 worn on a wearable device, which may be an edge or IoT device. This data is used to detect falls in a trained fall detection machine learning model by a trained fall detection algorithm, i.e., a trained fall detection computer program (see "Fall Detection ML Algorithm (ML or Analytics)" 300, "Detect Fall?" 310 in FIG. 33). "Y" means "yes" and "N" means "no".

[0071] The fall detection algorithm, i.e., the fall detection computer program, and the fall detection machine learning model are trained by training data collected by user 1, who may be, for example, a trainer or other user, such as a motorcyclist and bicycle rider, who wears the device and collects relevant fall data to train the fall detection algorithm, i.e., the fall detection computer program, and the fall detection machine learning model. The data may also be labeled as the data is being collected, or may be labeled later offline, using either a manual or automatic approach. If a fall is predicted by the fall detection algorithm, i.e., the fall detection computer program, another algorithm, i.e., another computer program, a fall type prediction algorithm, i.e., a fall type prediction computer program, in the "Fall Type Prediction Machine Learning Module" is triggered, which predicts the type of the detected fall. The "Fall Type Prediction Algorithm (ML or Analytics)" 320 and the "Prediction of Fall Type of Detected Fall" are included in the fall characterization according to the present disclosure, and the "Fall Characterization Machine Learning Unit" includes the "Fall Type Prediction Machine Learning Module" as described herein according to the present disclosure.

[0072] The exact fall type classification, i.e., fall type characterization, i.e., both are included in the fall characterization according to the present disclosure, will depend on the particular user case, for example, one method is to classify, i.e., characterize the classification of fall basing based on the fall angle, such as forward fall, backward fall, side fall (left / right), or a fall in place without significant leaning towards any side or angle, or a fall based on a slip.

[0073] In parallel, a fall severity prediction algorithm 330, i.e., a fall severity prediction computer program that may or may not be included in a "Fall Severity Prediction Machine Learning Module", is used, which may or may not be based on machine learning (see Fall Severity Prediction Algorithm (ML or Analytics) 330 in FIG. 3). This fall severity prediction algorithm 330, i.e., a fall severity prediction computer program, predicts and / or estimates the severity of a fall based on fall acceleration, angle, and other parameters (position, tilt, etc.). The fall severity prediction algorithm 330 is included in the "Fall Severity Prediction Machine Learning Module" according to the present disclosure.

[0074] As indicated above, Fall Detection 300, Fall Type Prediction 320, and Fall Severity Prediction 330 may or may not be ML based. Non-ML based methods may be based on domain knowledge, for example.

[0075] If both of these characteristics, i.e. "Fall Type Prediction" and "Fall Severity Prediction", are not predicted with high certainty (or confidence) (see "Both Predicted with High Confidence?" 340 in FIG. 3, "Y" means "Yes" and "N" means "No") and "Is Fall Detected?" in FIG. 3, then a predefined default actuation method, i.e. actuation, may be signaled to be triggered (see "Default Actuation" in FIG. 3). Otherwise, if both of the characteristics are indeed triggered with high confidence, the time to impact is calculated (see "Predict Time to Impact" 350 in FIG. 3), which finally leads to an actuation selection algorithm 360, i.e. actuation selection computer program, which extracts from the already generated action plan for actuation based on the predicted and calculated characteristics such as fall type, fall severity, and predicted time to impact. Finally, the selected actuation is signaled to be performed (370), the selected actuation being appropriate for that particular user for that particular crash or fall incident.

[0076] According to some aspects, as described herein, a computer-implemented method according to the present disclosure is disclosed, wherein the “fall characterization machine learning unit” comprises a “fall severity prediction machine learning module” that enables prediction of the severity of a detected fall during said implementation of a trained machine learning model.

[0077] Further, with reference to FIG. 4 , the present disclosure also relates to a control unit arrangement 700 for user fall assessment, the control unit arrangement 700 adapted to at least control the implementation of a trained machine learning model as described herein and / or the enabling of actuations according to an action plan of the trained machine learning model as described herein.

[0078] FIG. 4 illustrates diagrammatically the components of a control unit 700 according to one embodiment with respect to several functional units.

[0079] The processing circuitry 710 may be provided using any combination of one or more suitable central processing units (CPUs), multiprocessors, microcontrollers, digital signal processors (DSPs), special purpose hardware accelerators, etc. capable of executing software instructions stored in a computer program product, such as in the form of a storage medium 730. The processing circuitry 710 may further be provided as at least one application specific integrated circuit (ASIC) or field programmable gate array (FPGA).

[0080] In particular, the processing circuitry 710 is configured to cause the control unit 700 to perform a set of operations or steps, which were discussed above in connection with the various radar transceivers and methods. For example, the storage medium 730 may store a set of operations, and the processing circuitry 710 may be configured to retrieve the set of operations from the storage medium 730 and cause the control unit 700 to perform the set of operations. The set of operations may be provided as a set of executable instructions. Thus, the processing circuitry 710 is configured to thereby perform the methods and operations disclosed herein.

[0081] The storage medium 730 may also comprise persistent storage, which may be, for example, any one or combination of magnetic memory, optical memory, solid state memory, or even remotely attached memory.

[0082] The control unit 700 may further comprise a communication interface 720 for communicating with at least one other unit. The communication interface 720 may thus comprise one or more transmitters and receivers with analog and digital components and a suitable number of ports for wired or wireless communication.

[0083] The processing circuitry 710 is adapted to control the general operation of the control unit 700, for example, by sending data and control signals to an external unit and storage medium 730, by receiving data and reports from an external unit, and by retrieving data and instructions from the storage medium 730. Other components and associated functions of the control unit 700 have been omitted so as not to obscure the concepts presented herein.

[0084] For all embodiments disclosed herein, it should be understood that Figure 5 discloses a general representation of a computer program 810 including computer readable instructions 820 on a computer readable medium 830. The computer program may be considered, according to some aspects, a computer program product.

[0085] The present disclosure also relates to a computer program 810 comprising computer readable instructions 820 for applying the computer-implemented methods as described herein, and to a computer readable medium 830 comprising said computer program 810.

[0086] The present disclosure also relates to a computer program 810 including computer readable instructions 820 for training a machine learning model as described herein, and a computer readable medium 830 including the computer program 810.

[0087] The present disclosure also relates to a computer program 810 comprising computer readable instructions 820 for a "fall characterization machine learning unit" as described herein, and a computer readable medium 830 comprising said computer program 810.

[0088] The present disclosure also relates to a device, e.g., a wearable device, for enabling implementation of a trained machine learning model for communication of information for user fall assessment, the trained machine learning model for enabling actuation according to an action plan of the trained machine learning model as described herein, the action plan being generated using a fall characterization machine learning unit, the fall characterization machine learning unit being trained in fall characterization, and the fall assessment including fall detection.

[0089] The device comprises a trained machine learning model trained on a user fall assessment and including an action plan, device sensors 2, 3, 6, 7 for detecting data, means for continuously acquiring data, means 700 for processing the data, means for generating information based on the acquired and processed data, and means for communicating the information.

[0090] The device comprises a computer program 810 and / or a computer readable medium 830 for applying the computer implemented method as described herein, for training machine learning models as described herein, and / or for a "fall characterization machine learning unit" as described herein, all as described herein.

[0091] According to some embodiments, the device is at least partially in the form of a vest 10 or similar garment. The device can also be configured with a control unit arrangement 700 and / or a bracelet or similar. In FIG. 1, a user 1 is shown listening to a sensor 2 included in a bracelet. In general, the present disclosure relates to a garment 10 comprising a control unit arrangement 700 as described herein, one or more sensors 2, 3, 6, 7 for detecting data as described herein, and at least one airbag 4, 5, 9.

[0092] The present disclosure also relates to a system for enabling implementation of a trained machine learning model for communication of information for user fall assessment, the trained machine learning model for enabling actuation according to an action plan of the trained machine learning model as described herein, the action plan being generated using a fall characterization machine learning unit, the fall characterization machine learning unit being trained in fall characterization, and the fall assessment including fall detection.

[0093] The system comprises a trained machine learning model trained on a user fall assessment and including an action plan, sensors 2, 3, 6, 7 for detecting data, means for continuously acquiring data, means 700 for processing the data, means for generating information based on the acquired and processed data, and means for communicating the information.

[0094] The system comprises a computer program 810 and / or a computer readable medium 830 for applying the computer-implemented method as described herein, for training a machine learning model as described herein, and / or for a "fall characterization machine learning unit" as described herein, all as described herein.

[0095] According to some aspects, the present disclosure relates to a computer-implemented method for fall assessment for a fall assessment environment, the computer-implemented method including implementing a trained machine learning model including a “fall characterization machine learning unit,” an action plan, and an actuation according to the action plan, the action plan being generated reflecting training of the machine learning model.

[0096] The actuation according to the action plan includes measures, such as protective measures, preventive measures, and / or warning measures, reflecting a fall assessment for the user, the fall assessment including fall detection and fall characterization. The computer-implemented method includes continuously acquiring data and continuously processing the data, the computer-implemented method includes continuously producing information based on the acquired data and the processed data, the computer-implemented method includes continuously communicating the information, the computer-implemented method includes a fall assessment for the user, the fall assessment including fall detection and fall characterization based on the continuous data and the continuous information.

[0097] The fall assessment environment comprises a trained machine learning model trained in fall assessment, a user 1, sensors 2, 3, 6, 7 for detecting data from the user 1's fall, means for continuously acquiring data, means 700 for processing the data, means for generating information based on the acquired and processed data, means for communicating the information, and means for actuation according to an action plan and based on the successive data and the successive information.

[0098] The trained machine learning model, during its implementation in the computer-implemented method for fall assessment, continuously enables the acquisition of data and the continuous processing of data, the continuous production and continuous communication of information, fall assessment for user 1, including fall detection and fall characterization, and actuation that follows an action plan and reflects the fall assessment.

[0099] The "fall characterization machine learning unit" includes a "fall type prediction machine learning module" which, during said implementation of the trained machine learning model, enables prediction of a fall type of a detected fall when a fall is detected. The computer-implemented method for fall assessment includes fall characterization of a detected fall, the fall characterization including prediction of a fall type, and action plans and actuations according to the predicted fall type.

[0100] According to some embodiments, there may be other sensors besides the 3D sensors mentioned above, such as 1D / 2D motion sensors as well as magnetometers, GPS, etc.

Claims

1. 1. A computer-implemented method for fall assessment for a fall assessment environment, the computer-implemented method comprising: implementing a trained machine learning model comprising a plan of action; and actuation according to the plan of action; The action plan is generated reflective of training of the machine learning model; The actuation according to the action plan includes protective measures and reflects a fall assessment for a user (1), the fall assessment including fall detection and fall characterization of the detected fall, and the computer-implemented method includes continuously acquiring data and continuously processing data, the continuously acquired data including data detected by sensors (2, 3, 6, 7) worn by the user (1).

2. 2. The computer-implemented method of claim 1, wherein the protective measures include physical measures, including then inflating an airbag (4, 5, 9).

3. The computer-implemented method of claim 1 or 2, wherein the actuation according to the action plan then includes measures including at least one of preventive measures and alarm measures.

4. 4. The computer-implemented method of claim 1, wherein the computer-implemented method includes a continuous production of information based on the acquired data and the processed data, the computer-implemented method includes a continuous communication of the information, and the computer-implemented method includes a fall assessment relating to a user (1), the fall assessment including fall detection and fall characterization based on continuous data and the continuous information.

5. The fall evaluation environment includes: the trained machine learning model trained in fall assessment; A user (1), sensors (2, 3, 6, 7) for detecting data from a fall of a user (1); means for continuously acquiring data; A means (700) for processing data; means for generating said information based on the acquired and processed data; means for communicating said information; means for said actuation according to said action plan and based on said sequential data and said sequential information; The trained machine learning model, during its implementation in the computer-implemented method for fall assessment, continuously acquiring said data and continuously processing said data; said continuous production and said continuous communication of said information; the fall assessment for the user (1), including the fall detection and the fall characterization; The computer-implemented method of claim 1 , further comprising: continuously enabling the actuation according to the action plan and reflective of the fall assessment.

6. 6. The computer-implemented method of claim 1, wherein the trained machine learning model continuously enables, during its implementation in the computer-implemented method for fall assessment, an estimation of a confidence in any prediction in the computer-implemented method, the computer-implemented method comprising: estimating a confidence in any prediction of a fall characterization; selecting an actuation in accordance with the action plan and the prediction if the confidence in the prediction is high; and selecting a default actuation in accordance with the action plan if the confidence in the prediction is not high.

7. The computer-implemented method of any one of claims 1 to 6, wherein the computer-implemented method comprises predicting a severity of the detected fall.

8. the trained machine learning model, during its implementation in the computer-implemented method for fall assessment, continuously enables an estimation of a time to impact of the detected fall; The computer-implemented method of any one of claims 1 to 7, wherein the computer-implemented method includes estimating a time to impact of the detected fall.

9. The computer-implemented method of claim 1 , wherein the fall characterization includes a prediction of a fall type, and the actuation is performed according to the action plan and the predicted fall type.

10. A computer program (810) comprising computer readable instructions (820) for applying the computer implemented method of any one of claims 1 to 9, and a computer readable medium (830) comprising said computer program (810).

11. A computer program (810) comprising computer readable instructions (820) for a trained machine learning model according to any one of claims 1 to 9, and a computer readable medium (830) comprising said computer program (810).

12. A control unit arrangement (700) for user fall assessment adapted to at least control the implementation of a trained machine learning model according to any one of claims 1 to 9 and / or the enabling of the actuation according to an action plan of the trained machine learning model.

13. A device, e.g. a wearable device, for a user fall assessment, for communication of information, for enabling the implementation of a trained machine learning model according to any one of claims 1 to 9 and for enabling actuation according to a behavioral plan of said trained machine learning model, comprising: The action plan is generated using a fall characterization machine learning unit, the fall characterization machine learning unit being trained on fall characterization; The fall assessment includes fall detection; The device comprises: the trained machine learning model trained on the user fall assessment and including the action plan; Device sensors (2, 3, 6, 7) for detecting data; means for continuously acquiring data; A means (700) for processing data; means for generating said information based on the acquired and processed data; and means for communicating said information; The device comprises a computer program (810) according to one or more of claims 10 and 11 and / or a computer readable medium (830) according to one or more of claims 10 and / or 11.

14. A system for enabling the implementation of a trained machine learning model according to any one of claims 1 to 9 for a user fall assessment, for communication of information, and for enabling actuation according to a behavioral plan of said trained machine learning model, comprising: The action plan is generated using a fall characterization machine learning unit, the fall characterization machine learning unit being trained on fall characterization; The fall assessment includes fall detection; The system comprises: the trained machine learning model trained on the user fall assessment and including the action plan; Sensors (2, 3, 6, 7) for detecting data; means for continuously acquiring data; A means (700) for processing data; means for generating said information based on the acquired and processed data; and means for communicating the information, the system comprising a computer program (810) according to one or more of claims 10 and 11, and / or a computer readable medium (830) according to one or more of claims 10 and 11.

15. A garment (10) comprising a control unit arrangement (700) according to claim 12, one or more sensors (2, 3, 6, 7) for detecting data, and at least one airbag (4, 5, 9).

Citation Information

Patent Citations

  • Fall detecting method and fall rescuing device

    JP2004252618A

  • Computerized automatic dynamic control system for motorcycles

    JP2006513075A

  • Fall detection using mobile devices

    JP2020531110A

  • Fall detection terminal and program

    JP2021108047A

  • Identifying and responding to vehicle internal conditions

    JP2021509876A