Drowsiness-estimating device and drowsiness-estimating method
View Patent ↗A drowsiness-estimating device capable of improving the precision of drowsiness estimation by eliminating the effect of individual differences. In the device, a recurrence required time-calculating part ( 103 ) calculates the recurrence required time, which is the time needed, after a detection time when an action is detected, for a drowsiness estimation parameter value acquired after the detection time to return to the value of the drowsiness estimation parameter acquired before the detection time. A drowsiness-estimating part ( 104 ) estimates the level of drowsiness of the drowsiness-estimation subject on the basis of the calculated recurrence required time. To be specific, the drowsiness-estimating part ( 104 ) maintains a drowsiness level-estimating table in which each of multiple time ranges is correlated with a possible drowsiness level, and specifies a possible drowsiness level that corresponds to the time range, among the multiple time ranges, with which the recurrence required time is associated.
1. A drowsiness estimating apparatus for estimating a drowsiness level of a subject of drowsiness estimation from among a plurality of drowsiness level candidates, the drowsiness estimating apparatus comprising:
an acquisition section that acquires a drowsiness estimation parameter at each of a plurality of acquisition timings;
a detecting section that detects an action of the subject of drowsiness estimation;
a time-calculating section that calculates a recurrence required time which is the time it takes, from a detection timing at which the action is detected, for a value of the drowsiness estimation parameter acquired after the detection timing to return to a value of the drowsiness estimation parameter acquired before the detection timing; and
an estimating section that estimates a drowsiness level of the subject of drowsiness estimation based on the calculated recurrence required time.
2. The drowsiness estimating apparatus according to claim 1 , wherein:
the acquisition section comprises a level calculation section that calculates a second drowsiness level as the drowsiness estimation parameter based on face expression feature information of the subject of drowsiness estimation; and
the time-calculating section calculates the recurrence required time only when the second drowsiness level is associated with a drowsiness level candidate other than the highest and lowest drowsiness levels among the plurality of the drowsiness level candidates.
3. The drowsiness-estimating apparatus according to claim 1 , wherein the estimation section holds a drowsiness level estimation table in which a plurality of time ranges are associated with the drowsiness level candidates, respectively, and the estimation section identifies a drowsiness level candidate corresponding to a time range to which the calculated recurrence required time belongs from among the plurality of time ranges.
4. The drowsiness-estimating apparatus according to claim 3 , further comprising:
an action-classifying section that classifies an action detected by the detection section into one of a plurality of classes based on an action classification parameter; and
a table-adjusting section that adjusts a time defining each of the time ranges of the drowsiness level estimation table in accordance with the class of classification.
5. The drowsiness-estimating apparatus according to claim 4 , wherein the action classification parameter is a size of the detected action.
6. The drowsiness-estimating apparatus according to claim 4 , wherein the action classification parameter is a frequency of occurrence of an event that evokes the detected action.
7. A drowsiness-estimating method that estimates a drowsiness level of a subject of drowsiness estimation from among a plurality of drowsiness level candidates, the drowsiness-estimating method comprising:
acquiring a drowsiness estimation parameter at each of a plurality of acquisition timings;
detecting an action of the subject of drowsiness estimation;
calculating a recurrence required time which is the time it takes, from a detection timing at which the action is detected, for a value of the drowsiness estimation parameter acquired after the detection timing to return to a value of the drowsiness estimation parameter acquired before the direction timing; and
estimating a drowsiness level of the subject of drowsiness estimation based on the calculated recurrence required time.