A modified Hidden Markov Model (HMM)-based state machine model for driving behavior recognition: Effectiveness of features using different sub-HMMs
DOI: 10.29007/g4sc
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of accurately recognizing driving behaviors, specifically lane changes, for Advanced Driving Assistance Systems (ADAS). While machine learning approaches are common, selecting optimal model configurations and input features remains difficult. The authors propose an improved Hidden Markov Model (HMM)-based state machine model that integrates a prefilter to quantize input variables and utilizes multiple sub-HMMs to handle different data types. The study aims to evaluate the effectiveness of various feature combinations by fusing different sub-HMMs and optimizing model parameters to enhance estimation accuracy, detection rates, and reduce false alarms. The methodology combines a three-state machine (representing lane change right, lane keeping, and lane change left) with an improved HMM. The HMM consists of four sub-models, each processing distinct input variables: time to collision (TTC), distances, velocities, and driving operational variables (e.g., steering angle, pedal position). A prefilter with five thresholds quantizes these inputs into observation sequences. To optimize the prefilter thresholds and the weights used to fuse sub-HMM probabilities, the authors employ the Non-Dominated Sorting Genetic Algorithm II (NSGA-II), a multi-objective optimization technique. The model was trained and tested using experimental data collected from six drivers in a driving simulator, involving 40-minute training sessions and 10-minute test sessions on a four-lane highway environment. The results evaluate eleven different combinations of sub-HMMs. The analysis reveals that the combination including driving operational variables (HMM V) achieved the highest number of top-performing metrics, though it exhibited imbalanced performance in specific detection and false alarm rates. Conversely, combinations HMM VIII, X, and XI demonstrated balanced performance across all metrics. When compared to a conventional HMM with default parameters, the proposed approach using combinations VIII, X, and XI outperformed the conventional model in most metrics. Notably, all high-performing combinations included driving operational variables, highlighting their critical role in accurate behavior recognition. The conventional HMM only outperformed the proposed method in specific cases (HMM III and V), where the proposed model’s detection rates were lower. The significance of this work lies in demonstrating that a modified HMM structure, combined with optimized prefilters and specific feature selections, can effectively recognize lane-changing behaviors. The findings underscore the importance of driving operational variables as inputs for such models. The study provides a framework for selecting effective feature combinations and optimizing hyperparameters in driving behavior estimation, offering a robust alternative to conventional HMMs for ADAS development. Future work may explore additional environmental features and increased prefilter thresholds to further refine model specifications.
Key finding
Sub-HMM combinations that include driving operational variables, particularly HMM VIII, X, and XI, yielded balanced performance and superior accuracy compared to conventional HMM models for lane-changing behavior recognition.
Methodology
simulation_modeling
Sample size: 6
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-28.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | unpaywall | — | — | 2 | 2026-06-04 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | — | — | — | 1 | 2026-05-28 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Theoretical Contribution: computational model