Driving context generation system for generating driving behavior description information

1. A driving context generation system comprising:

a discretizer configured to discretize at least one driving behavioral data sequence representing time-series driving behaviors of a vehicle into segmented behavior-data sequences, each of the segmented behavior-data sequences corresponding to one of sequential driving situations of the vehicle;

a behavior description generator configured to:

extract a feature portion from at least one of the segmented behavior-data sequences corresponding to at least one target driving situation included in the sequential driving situations; and

generate behavior-description information indicative of a description of a driver's behavior associated with the at least one target driving situation;

an event description generator configured to:

sequentially detect traffic events in at least one environmental data sequence indicative of time-series driving environments of the vehicle, each of the detected traffic events having a potential to impact on driver's driving behaviors of the vehicle; and

generate event-description information including a description indicative of each of the detected traffic events; and

a context description generator configured to:

extract, from the sequentially detected traffic events, one target traffic event that is detected immediately adjacent to the feature portion; and

generate context information indicative of a context between the behavior-description information generated by the behavior description generator and the event-description information about the target traffic event generated by the event description generator.

2. The driving context generation system according to claim 1 , wherein the context description generator is configured to:

obtain information indicative of a temporal relationship between the feature portion and the target traffic event; and

combine, based on the temporal relationship, the behavior description information with the event-description information, thus generating the context information.

3. The driving context generation system according to claim 1 , wherein:

the behavior description generator comprises:

a feature-quantity distribution generator configured to generate, for each of the driving situations:

a feature-quantity distribution of appearance frequencies of feature quantities included in a corresponding one of the segmented behavior-data sequences; and

a feature-quantity distribution of appearance frequencies of feature quantities included in the at least one environmental data sequence, thus obtaining a group of the feature-quantity distributions for each of the driving situations;

a topic proportion calculator configured to:

store therein a group of driving topics each representing a corresponding basic driving situation that frequently appears while the vehicle is travelling, information about the driving topics being used to express the group of the feature-quantity distributions for each of the driving situations as a combination of at least ones of the driving topics;

select ones of the driving topics for each of the driving situations; and

calculate, as a topic proportion for each of the driving situations, percentages of the selected driving topics with respect to the whole of the selected driving topics such that the calculated percentages of the selected driving topics for each of the driving situations expresses the group of the feature-quantity distributions included in a corresponding one of the driving situations;

a topic description storage configured to store therein driver behavioral descriptions defined to correlate with the respective driving topics; and

a behavioral description extracting unit configured to:

extract the feature portion from the sequential driving situations;

extract, as a noticeable driving topic, one of the driving topics that is noticeable at the at least one target driving situation according to the calculated topic proportion for the at least one target driving situation; and

extract one of the driver behavioral descriptions correlating with the noticeable driving topic as the behavior-description information.

4. The driving context generation system according to claim 3 , wherein the behavior description generator is configured to:

calculate a divergence between the topic proportions of each pair of temporally adjacent driving situations in the sequential driving situations; and

extract, as the feature portion, a specific pair of temporally adjacent first and second driving situations when the divergence of the topic proportions of the specific pair of temporally adjacent first and second driving situations is greater than a predetermined threshold.

5. The driving context generation system according to claim 4 , wherein the behavior description generator is configured to:

extract, as the noticeable driving topic, a driving topic whose corresponding percentage during a duration of the temporally adjacent first and second driving situations changes most significantly in the percentages of all the driving topics during the duration of the temporally adjacent first and second driving situations.

6. The driving context generation system according to claim 4 , wherein the behavior description generator is configured to:

repeatedly extract the feature portion;

calculate an average topic proportion based on all driving situations belonging to a stable range defined as a range between the repeatedly extracted feature portions; and

extract, from the average topic proportion, a driving topic having a maximum percentage in the average topic proportion as a second noticeable driving topic.

7. The driving context generation system according to claim 5 , wherein the behavior description generator is configured to:

repeatedly extract the feature portion;

calculate an average topic proportion based on all driving situations belonging to a stable range defined as a range between the repeatedly extracted feature portions; and

extract, from the average topic proportion, a driving topic having a maximum percentage in the average topic proportion as a second noticeable driving topic.

8. A computer program product for a driving context generation system, the computer program product comprising:

a non-transitory computer-readable storage medium; and

a set of computer program instructions embedded in the computer-readable storage medium, the instructions causing a computer to carry out:

a first step of discretizing at least one driving behavioral data sequence representing time-series driving behaviors of a vehicle into segmented behavior-data sequences, each of the segmented behavior-data sequences corresponding to one of sequential driving situations of the vehicle;

a second step of extracting a feature portion from at least one of the segmented behavior-data sequences corresponding to at least one target driving situation included in the sequential driving situations;

a third step of generating behavior-description information indicative of a description of a driver's behavior associated with the at least one target driving situation;

a fourth step of sequentially detecting traffic events in at least one environmental data sequence indicative of time-series driving environments of the vehicle, each of the detected traffic events having a potential to impact on driver's driving behaviors of the vehicle;

a fifth step of generating event-description information including a description indicative of each of the detected traffic events;

a sixth step of extracting, from the sequentially detected traffic events, one target traffic event that is detected immediately adjacent to the feature portion; and

a seventh step of generating context information indicative of a context between the behavior-description information generated by the third step and the event-description information about the target traffic event generated by the fifth step.

9. A driving context generation method comprising:

a first step of discretizing at least one driving behavioral data sequence representing time-series driving behaviors of a vehicle into segmented behavior-data sequences, each of the segmented behavior-data sequences corresponding to one of sequential driving situations of the vehicle;

a second step of extracting a feature portion from at least one of the segmented behavior-data sequences corresponding to at least one target driving situation included in the sequential driving situations;

a third step of generating behavior-description information indicative of a description of a driver's behavior associated with the at least one target driving situation;

a fourth step of sequentially detecting traffic events in at least one environmental data sequence indicative of time-series driving environments of the vehicle, each of the detected traffic events having a potential to impact on driver's driving behaviors of the vehicle;

a fifth step of generating event-description information including a description indicative of each of the detected traffic events;

a sixth step of extracting, from the sequentially detected traffic events, one target traffic event that is detected immediately adjacent to the feature portion; and

a seventh step of generating context information indicative of a context between the behavior-description information generated by the third step and the event-description information about the target traffic event generated by the fifth step.