Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unsupervised Work Behavior Pattern Extraction Based on Hierarchical Probabilistic Model

Issei Saito, Tomoaki Nakamura, Toshiyuki Hatta, Wataru Fujita, Shintaro Watanabe, Shotaro Miwa

TL;DR

This work tackles unsupervised analysis of industrial worker behavior by introducing a two-layer hierarchical probabilistic model that couples a bottom GP-HSMM with a top HSMM. Through mutual learning between layers, it improves segmentation of both motion elements and unit motions without labeled data, outperforming baselines on real production data. The motion-element unigram emission (ME-U) within the HSMM yielded the best unit-motion segmentation, demonstrating robustness to natural variability in worker actions. The approach enables rapid, automated feedback for process improvement in high-mix, low-volume manufacturing, with potential extensions to nonparametric class counts and real-time deployment.

Abstract

Evolving consumer demands and market trends have led to businesses increasingly embracing a production approach that prioritizes flexibility and customization. Consequently, factory workers must engage in tasks that are more complex than before. Thus, productivity depends on each worker's skills in assembling products. Therefore, analyzing the behavior of a worker is crucial for work improvement. However, manual analysis is time consuming and does not provide quick and accurate feedback. Machine learning have been attempted to automate the analyses; however, most of these methods need several labels for training. To this end, we extend the Gaussian process hidden semi-Markov model (GP-HSMM), to enable the rapid and automated analysis of worker behavior without pre-training. The model does not require labeled data and can automatically and accurately segment continuous motions into motion classes. The proposed model is a probabilistic model that hierarchically connects GP-HSMM and HSMM, enabling the extraction of behavioral patterns with different granularities. Furthermore, it mutually infers the parameters between the GP-HSMM and HSMM, resulting in accurate motion pattern extraction. We applied the proposed method to motion data in which workers assembled products at an actual production site. The accuracy of behavior pattern extraction was evaluated using normalized Levenshtein distance (NLD). The smaller the value of NLD, the more accurate is the pattern extraction. The NLD of motion patterns captured by GP-HSMM and HSMM layers in our proposed method was 0.50 and 0.33, respectively, which are the smallest compared to that of the baseline methods.

Unsupervised Work Behavior Pattern Extraction Based on Hierarchical Probabilistic Model

TL;DR

This work tackles unsupervised analysis of industrial worker behavior by introducing a two-layer hierarchical probabilistic model that couples a bottom GP-HSMM with a top HSMM. Through mutual learning between layers, it improves segmentation of both motion elements and unit motions without labeled data, outperforming baselines on real production data. The motion-element unigram emission (ME-U) within the HSMM yielded the best unit-motion segmentation, demonstrating robustness to natural variability in worker actions. The approach enables rapid, automated feedback for process improvement in high-mix, low-volume manufacturing, with potential extensions to nonparametric class counts and real-time deployment.

Abstract

Evolving consumer demands and market trends have led to businesses increasingly embracing a production approach that prioritizes flexibility and customization. Consequently, factory workers must engage in tasks that are more complex than before. Thus, productivity depends on each worker's skills in assembling products. Therefore, analyzing the behavior of a worker is crucial for work improvement. However, manual analysis is time consuming and does not provide quick and accurate feedback. Machine learning have been attempted to automate the analyses; however, most of these methods need several labels for training. To this end, we extend the Gaussian process hidden semi-Markov model (GP-HSMM), to enable the rapid and automated analysis of worker behavior without pre-training. The model does not require labeled data and can automatically and accurately segment continuous motions into motion classes. The proposed model is a probabilistic model that hierarchically connects GP-HSMM and HSMM, enabling the extraction of behavioral patterns with different granularities. Furthermore, it mutually infers the parameters between the GP-HSMM and HSMM, resulting in accurate motion pattern extraction. We applied the proposed method to motion data in which workers assembled products at an actual production site. The accuracy of behavior pattern extraction was evaluated using normalized Levenshtein distance (NLD). The smaller the value of NLD, the more accurate is the pattern extraction. The NLD of motion patterns captured by GP-HSMM and HSMM layers in our proposed method was 0.50 and 0.33, respectively, which are the smallest compared to that of the baseline methods.
Paper Structure (13 sections, 10 equations, 5 figures, 4 tables, 2 algorithms)

This paper contains 13 sections, 10 equations, 5 figures, 4 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related work
  3. Proposed model
  4. Generative process
  5. Gaussian Process
  6. Parameter Inference
  7. Inference Algorithm
  8. Emission Probability of HSMM and Prior Probability of GP-HSMM
  9. Experiment
  10. Experimental Setup
  11. Segmentatation of motion elements
  12. Segmentation of unit motions
  13. Conclusion

Figures (5)

  • Figure 1: Overview
  • Figure 2: Graphical model of GP-HSMM-BA
  • Figure 3: Work scenario
  • Figure 4: Normalized Levenshtein distance of 10 trials
  • Figure 5: Visualization of unit motion segmentation. Left: ground truth, Middle: segments estimated by GP-HSMM+ME-U HSMM, and Right: segments estimated by GP-HSMM+ME-B HSMM.