PHINN-EEG: Topological Time-Series Analysis of Dream-State EEG -- Dynamic Betti Curves for Dream Content Classification and Topology-Conditioned Neural Signal Synthesis

Ren Takahashi, Emre Yusuf, Jayabrata Bhaduri|July 10, 2026arXiv

Key Takeaway

Topological features derived from phase-space geometry of EEG outperform traditional spectral features for dream detection, suggesting that the shape of neural activity patterns matters more than their energy content.

Summary

This paper introduces PHINN-EEG, a topological data analysis framework for detecting dream states from EEG signals. Instead of traditional power spectrum features, it uses persistent homology to extract geometric patterns (Dynamic Betti Curves) from neural activity, aiming to improve dream detection accuracy from 70% to 82-90% AUC.

evaluation data reasoning

Key Terms

persistent-homology betti-number takens-delay-embedding vietoris-rips-filtration rectified-flow