System and method for event prevention and prediction

1. A computer-implemented method, comprising:

based on information of a subject positioned on an object in an environment, generating a data stream;

for the data stream, extracting features associated with a movement of the subject with respect to the object or the environment, wherein the movement is represented by spatio-temporal features extracted from sensors;

generating a prediction associated with a likelihood of the movement based on the extracted features, and a risk profile of the movement based on a plurality of fall risk factors; and

applying the prediction and the risk profile to a rule base to perform an action, wherein the rule base comprises determining whether the prediction exceeds a first threshold and whether the risk profile exceeds a second threshold, and based on the determining, performing the action.

2. The computer-implemented method of claim 1 , wherein the information is sensed by at least one of a camera sensor and a motion sensor.

3. The computer-implemented method of claim 2 , wherein the camera sensor comprises a depth sensor or a thermal sensor.

4. The computer-implemented method of claim 1 , wherein the extracting the features comprises applying the dense trajectories to extract the features or Convolutional Neural Networks (CNN) to learn the spatio-temporal features from a sequence of frames in the data stream, applying an automatic learning system to learn the features, and encoding sequences of feature vectors based on the learned features and the extracted features.

5. The computer-implemented method of claim 1 , wherein the generating the prediction comprises providing, over a time horizon, the extracted features to a recurrent neural network or a convolutional neural network to generate an output that is provided to the dense layer, an output of which is fed to a softmax layer to generate the prediction.

6. The computer-implemented method of claim 1 , wherein the risk profile is generated by receiving the plurality of fall risk factors that comprise factors specific to the subject, factors specific to an environment associated with the subject, and factors associated with behavioral routines associated with the environment and the subject, over time.

7. The computer-implemented method of claim 1 , wherein the action comprises one or more of generating a signal for the subject, controlling a state of the object, and providing a command to an external resource to perform the action.

8. The computer-implemented method of claim 1 , wherein the subject is a person, the object is a bed or a chair, and the movement is the person falling from the bed or the chair.

9. A non-transitory computer readable medium having a storage that stores instructions, the instructions executed by a processor, the instructions comprising:

based on information of a subject positioned on an object in an environment, generating a data stream;

for the data stream, extracting features associated with a movement of the subject with respect to the object or the environment, wherein the movement is represented by spatio-temporal visual features;

generating a prediction associated with a likelihood of the movement based on the extracted features, and a risk profile of the movement based on a plurality of fall risk factors; and

applying the prediction and the risk profile to a rule base to perform an action, wherein the rule base comprises determining whether the prediction exceeds a first threshold and whether the risk profile exceeds a second threshold, and based on the determining, performing the action.

10. The non-transitory computer readable medium of claim 9 , wherein the information is sensed by at least one of a camera sensor and/or a motion sensor, and the camera sensor comprises a depth sensor or a thermal sensor.

11. The non-transitory computer readable medium of claim 9 , wherein the extracting the features comprises applying the dense trajectories or Convolutional Neural Networks (CNN) to learn the spatio-temporal features to extract the features from a sequence of frames in the data stream, applying an automatic learning system to learn the features, and encoding sequences of feature vectors based on the learned features and the extracted features.

12. The non-transitory computer readable medium of claim 9 , wherein the generating the prediction comprises providing, over a time horizon, the extracted features to a recurrent neural network or a convolutional neural network to generate an output that is provided to the dense layer, an output of which is fed to a softmax layer to generate the prediction.

13. The non-transitory computer readable medium of claim 9 , wherein the risk profile is generated by receiving the plurality of fall risk factors that comprise factors specific to the subject, factors specific to an environment associated with the subject, and factors associated with behavioral routines associated with the environment and the subject, over time.

14. The non-transitory computer readable medium of claim 9 , wherein the action comprises one or more of generating a signal for the subject, controlling a state of the object, and providing a command to an external resource to perform the action.

15. The non-transitory computer readable medium of claim 9 , wherein the subject is a person, the object is a bed or a chair, and the movement is the subject falling from the bed or the chair.

16. A processor capable of processing a request, the processor configured to perform the operations of:

based on information of a subject positioned on a bed or chair in an environment, generating a data stream;

for the data stream, extracting features associated with a fall of the subject with respect to the bed or chair, or the environment, wherein the fall is represented by spatio-temporal visual features;

generating a prediction associated with a likelihood of the fall based on the extracted features, and a risk profile of the fall based on a plurality of fall risk factors; and

applying the prediction and the risk profile to a rule base to perform an action, wherein the rule base comprises determining whether the prediction exceeds a first threshold and whether the risk profile exceeds a second threshold, and based on the determining, performing the action.

17. The processor of claim 16 , wherein the extracting the features comprises applying the dense trajectories to extract the features or Convolutional Neural Networks (CNN) to learn the spatio-temporal features from a sequence of frames in the data stream, applying an automatic learning system to learn the features, and encoding sequences of feature vectors based on the learned features and the extracted features, the generating the prediction comprises providing, over a time horizon, the extracted features to a recurrent neural network or a deep neural network to generate an output that is provided to the dense layer, an output of which is fed to a softmax layer to generate the prediction, and the risk profile is generated by receiving the plurality of fall risk factors that comprise factors specific to the subject, factors specific to an environment associated with the subject, and factors associated with behavioral routines associated with the environment and the subject, over time.

18. The processor of claim 16 , wherein the action comprises one or more of generating a signal for the subject, controlling a state of the bed or chair, and providing a command to an external resource to perform the action.