Task estimation method and information processing apparatus

1. A task estimation method comprising:

acquiring, by a computer, first measurement information measured by a first device put on a task performer;

detecting a predetermined first operation in the first measurement information;

extracting first feature vectors from a first operation section in which the predetermined first operation has been detected;

dividing the extracted first feature vectors into first clusters by first division methods;

obtaining first ratios at which the extracted first feature vectors are classified into the respective first clusters in a predetermined time;

performing learning by using the first ratios as an input and using, as an output label, information indicating whether or not the task performer performs a predetermined task in the predetermined time;

generating a task estimation model by determining weighting for the first clusters based on a result of the learning; and

estimating, by using the task estimation model, whether or not the task performer is in a process of performing the predetermined task.

2. The task estimation method according to claim 1 , further comprising:

obtaining task estimation models by generating the task estimation model for each of tasks to be performed by the task performer; and

estimating a task performed by the task performer among the tasks by using the task estimation models.

3. The task estimation method according to claim 1 , further comprising:

acquiring second measurement information measured by a second device put on the task performer;

detecting a predetermined second operation in the second measurement information;

extracting second feature vectors from a second operation section in which the predetermined second operation has been detected;

dividing the extracted second feature vectors into second clusters by second division methods;

obtaining second ratios at which the extracted second feature vectors are classified into the respective second clusters in the predetermined time; and

performing the learning by further using the second ratios as the input.

4. The task estimation method according to claim 1 , further comprising:

obtaining task estimation models by generating the task estimation model for each of the first division methods;

selecting one task estimation model from the task estimation models based on estimation accuracies obtained from the task estimation models; and

estimating, by using the selected one task estimation model, whether or not the task performer is in a process of performing the predetermined task.

5. The task estimation method according to claim 3 , further comprising:

performing the learning, for each of combinations of the first division methods and the second division methods, by using the first ratios and the second ratios as the input;

obtaining task estimation models by generating the task estimation model for each of the combinations by determining weighting for the first clusters and the second clusters based on the result of the learning;

selecting one task estimation model from the task estimation models based on estimation accuracies obtained from the task estimation models; and

estimating, by using the selected one task estimation model, whether or not the task performer is in a process of performing the predetermined task.

6. An information processing apparatus comprising:

a processor configured to:

acquire first measurement information measured by a first device put on a task performer;

detect a predetermined first operation in the first measurement information;

extract first feature vectors from a first operation section in which the predetermined first operation has been detected;

divide the extracted first feature vectors into first clusters by first division methods;

obtain first ratios at which the extracted first feature vectors are classified into the respective first clusters in a predetermined time;

perform learning by using the first ratios as an input and using, as an output label, information indicating whether or not the task performer performs a predetermined task in the predetermined time;

generate a task estimation model by determining weighting for the first clusters based on a result of the learning; and

estimate, by using the task estimation model, whether or not the task performer is in a process of performing the predetermined task.

7. A non-transitory computer-readable recording medium having stored therein a program that causes a computer to execute a process, the process comprising:

acquiring first measurement information measured by a first device put on a task performer;

detecting a predetermined first operation in the first measurement information;

extracting first feature vectors from a first operation section in which the predetermined first operation has been detected;

dividing the extracted first feature vectors into first dusters by first division methods;

obtaining first ratios at which the extracted first feature vectors are classified into the respective first clusters in a predetermined time;

performing learning by using the first ratios as an input and using as an output label, information indicating whether or not the task performer performs a predetermined task in the predetermined time;

generating a task estimation model by determining weighting for he first dusters based on a result of the learning; and

estimating, by using the task estimation model, whether or not the task performer is in a process of performing the predetermined task.