Adaptive Temporal Dynamics for Personalized Emotion Recognition: A Liquid Neural Network Approach
Anindya Bhattacharjee, Nittya Ananda Biswas, K. A. Shahriar, Adib Rahman · 2026
Read Adaptive Temporal Dynamics for Personalized Emotion Recognition: A Liquid Neural Network Approach by Anindya Bhattacharjee, Nittya Ananda Biswas, K. A. Shahriar, Adib Rahman (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-06997.txt.
Description
Anindya Bhattacharjee and three co-authors apply liquid neural networks, whose neurons have learnable time constants, to emotion recognition from EEG signals. Body signals are non-steady, noisy, and highly person-dependent, making emotion recognition hard. Their multi-signal framework combines neural feature pulling, liquid networks, and attention-guided fusion to model time EEG dynamics alongside peripheral body signals and personality features.
Dedicated sub-nets process EEG and side signals, and a shared encoder-based fusion module learns sorting pictures before classification. On the PhyMER data set across seven emotion classes, subject-dependent tests reached about 95 percent accuracy, beating before reported results. Time attention analysis and charts back the findings, and stats analysis confirms the network self-organizes into fast and slow neuron groups that tune their time constants to the task.
Catalog record
| CREATOR | Anindya Bhattacharjee, Nittya Ananda Biswas, K. A. Shahriar, Adib Rahman |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-06997