Driver availability detection device and driver availability detection method
A driver availability detection device includes an information acquisition unit to acquire information related to an occupant and a threshold setting unit to set an abnormal-state determination threshold for estimating the abnormal state of the occupant on a basis of an abnormal state score obtained by inputting the information related to the occupant acquired by the information acquisition unit to a machine learning model within a first threshold setting time.
1. A driver availability detection device using a machine learning model, the driver availability detection device comprising:
processing circuitry
to acquire information related to the occupant,
to set an abnormal-state determination threshold that is adapted for the occupant, on a basis of an abnormal state score obtained by inputting information related to the occupant to the machine learning model that is trained using facial features, vehicle operational features, and biological features of a plurality of test occupants, within a first threshold setting time, and
to estimate an abnormal state of the occupant based on the abnormal-state determination threshold that is adapted for the occupant.
2. The driver availability detection device according to claim 1 , wherein the processing circuitry obtains one or more abnormal state scores and sets a maximum value of the one or more abnormal state scores as the abnormal-state determination threshold.
3. The driver availability detection device according to claim 1 , wherein the processing circuitry estimates that the occupant is in an abnormal state in a case where the abnormal state score obtained by inputting the information related to the occupant to the machine learning model is larger than the abnormal-state determination threshold.
4. The driver availability detection device according to claim 3 , wherein the processing circuitry individually authenticates the occupant and acquire personal identification information for identifying the occupant,
stores the abnormal-state determination threshold, which has been set, in association with the personal identification information, and
estimates the abnormal state of the occupant when acquiring the personal identification information and storing the abnormal-state determination threshold in association with the personal identification information.
5. The driver availability detection device according to claim 1 , wherein there are a plurality of the machine learning models,
the processing circuitry sets a model-specific threshold for each of the machine learning models on a basis of an abnormal state score obtained by the inputting information related to the occupant within the first threshold setting time, and
determines, after setting the model-specific threshold, whether or not an abnormal state score obtained by inputting information related to the occupant within a second threshold setting time is larger than the model-specific threshold for each of the machine learning models, and
sets a number of the machine learning models in which it is determined that the abnormal state score is larger than the model-specific threshold as the abnormal-state determination threshold.
6. The driver availability detection device according to claim 5 , wherein the processing circuitry calculates a number of machine learning models in which an abnormal state score obtained by inputting information related to the occupant is larger than a model-specific threshold, among the machine learning models, and
estimates that the occupant is in the abnormal state in a case where the number of the machine learning models is larger than the abnormal-state determination threshold.
7. The driver availability detection device according to claim 1 , wherein the processing circuitry acquires attribute information related to an attribute of the occupant,
selects one or more related machine learning models related to the attribute of the occupant among the machine learning models on a basis of the attribute information,
sets, for each of the one or more related machine learning models, a related-model-specific threshold on a basis of an abnormal state score obtained by inputting information related to the occupant within the first threshold setting time,
determines, after setting the related-model-specific threshold, whether or not an abnormal state score obtained by inputting information related to the occupant within a third threshold setting time is larger than the related-model-specific threshold for each of the one or more related machine learning models, and
sets a number of the related machine learning models in which it is determined that the abnormal state score is larger than the related-model-specific threshold as the abnormal-state determination threshold.
8. The driver availability detection device according to claim 7 , wherein the attribute of the occupant is an age of the occupant.
9. The driver availability detection device according to claim 7 , wherein the processing circuitry counts a number of related machine learning models in which an abnormal state score obtained by inputting information related to the occupant to the related machine learning model is larger than a related-model-specific threshold, among the machine learning models, and
estimates that the occupant is in an abnormal state in a case where the number of the related machine learning models is larger than the abnormal-state determination.
10. The driver availability detection device according to claim 1 , wherein the processing circuitry calculates a feature amount related to the occupant on a basis of information related to the occupant, and
sets the abnormal-state determination threshold on a basis of an abnormal state score obtained by inputting the feature amount to the machine learning model within the first threshold setting time.
11. A driver availability detection method using a machine learning model, the driver availability detection method comprising:
acquiring information related to the occupant; and
setting an abnormal-state determination threshold that is adapted for the occupant, on a basis of an abnormal state score obtained by inputting information related to the occupant to the machine learning model that is trained using facial features, vehicle operational features, and biological features of a plurality of test occupants, within a first threshold setting time; and
estimating an abnormal state of the occupant based on the abnormal-state determination threshold that is adapted for the occupant.