TOKYO, February 13, 2019 - Mitsubishi Electric Corporation (TOKYO: 6503) announced today that it has developed a unique behavioral-analysis artificial intelligence (AI) using the company's Maisart®* AI technology. Even without prior machine learning, the new technology can detect slight differences in human movements that people have difficulty noticing, which can be useful for analyzing human behavior in various fields, such as analyzing an assembly-line worker's motions to help eliminate unnecessary motions and thereby improve productivity.

* Mitsubishi Electric's AI creates the State-of-the-ART in Technology

Behavioral-analysis AI flowchart

Main Features

1)
Achieves fast processing without prior machine learning
- Processes behavioral analysis at high speed, within a few seconds or few minutes, less than 1/20th of the time required by the company's conventional AI method.

The new technology analyzes human movements immediately after collecting required measurement data, focusing on similarities in repeated movements. The technology is easy to apply at work sites because, unlike conventional behavior-analysis AI, it doesn't require machine learning involving huge amounts of teaching data that must be introduced manually. Behavioral analysis can be performed at high speed, within just a few seconds or minutes, which is less than 1/20th of the time required for the company's conventional method. Analysis can be performed quickly at work sites to provide rapid feedback on improving workers' efficiency.
2)
Detects slight differences in each person's motions to identify unnecessary motions
- Using position data to measure human movements, it estimates the boundaries between motions (operating elements), determines standard motion patterns for each person, and then detects deviations from these standard patterns, such as slightly different or unnecessary motions.
- On assembly lines, it can be used as a tool to help workers master optimal motions and thereby raise efficiency for improved productivity.

When analyzing assembly work performed in factories, the technology uses sensors to measure the three-dimensional positioning of both hands of a worker. This data makes it possible to detect non-standard motions such as slight procedural differences or unnecessary motions. In the beginning, the AI pays attention to motions repeated in a given order, such as attaching a part and screwing it into place. It divides the measurement data equally as the initial value and temporarily sets boundaries for each motion. Next, it extracts the waveform for each motion and compares it with the measurement data to update and determine the motion boundaries. Estimated motions are aligned to automatically determine standard motion patterns. Finally, by comparing the motion pattern extracted from measurement data with the standard motion pattern, it can detect non-standard motions.

Note that the releases are accurate at the time of publication but may be subject to change without notice.