Work analysis apparatus for analyzing work including series of actions performed by working subject

A work analysis apparatus for analyzing a work including a series of actions performed by a working subject, the work analysis apparatus comprising processing circuitry configured as:

• a sensor data input device that obtains a plurality of sensor data sequences, each of the sensor data sequences indicating time-series sensor values generated by measuring a work of the working subject using a sensor, the sensor data sequences being generated corresponding to repetitions of the work when the working subject repeatedly performs the work a plurality of times;
• a class data generator that determines a plurality of intervals obtained by temporally dividing each of the sensor data sequences based on the sensor values included in the sensor data sequences, determines classes of the intervals, each of the classes indicating a type of temporal variations in the sensor values included in one of the intervals, the classes being identified by integer symbols, and generates a plurality of first class data sequences for each of the sensor data sequences, each of the first class data sequences indicating the intervals and the classes of the sensor data sequence;
• a class data linker that associates the intervals having an identical class and corresponding to each other, with each other, among the plurality of first class data sequences, based on the plurality of first class data sequences;
• and
• a determiner that calculates characteristic values of the intervals associated with each other by the class data linker, and the determiner is further configured to determine, based on the characteristic values of the intervals, whether the actions of the working subject in the intervals should be improved,
• wherein the class data generator generates a plurality of standard patterns based on the plurality of sensor data sequences, each of the standard patterns indicating standard temporal variations in the sensor values included in each of the intervals, the standard patterns corresponding to the plurality of classes, respectively,
• wherein the class data generator generates the plurality of first class data sequences using the plurality of standard patterns,
• wherein the class data generator comprises:
• a first classifier that generates initial values of a plurality of second class data sequences for each of the sensor data sequences, each of the second class data sequences indicating the intervals and the classes of the sensor data sequence,
• a standard pattern generator that generates the plurality of standard patterns based on the plurality of sensor data sequences and the plurality of second class data sequences,
• a second classifier that generates a plurality of third class data sequences for each of the sensor data sequences using the plurality of standard patterns, each of the third class data sequences indicating the intervals and the classes of the sensor data sequence,
• a class data evaluator that updates the second class data sequences with the third class data sequences when the third class data sequences do not satisfy a predetermined evaluation criterion, and generates the third class data sequences as the first class data sequences when the third class data sequences satisfy the evaluation criterion, and
• wherein the class data generator repeats generating the standard patterns by the standard pattern generator, generating the third class data sequences by the second classifier, and updating the second class data sequences by the class data evaluator, until the third class data sequences satisfy the evaluation criterion.