# PhysioWave-T aims to cut biosignal model costs

> Researchers introduce wavelet-based models for EEG, ECG and muscle signals, reporting lower compute costs and improved results on selected tests.

Oossa · 2026-11-09 · https://oossa.com/en/physiowave-t-aims-to-cut-biosignal-model-costs

Event date: 2026-11-06

Researchers introduce PhysioWave-T, a family of AI models for EEG, ECG and sEMG, or muscle signals. The approach combines wavelet signal scales before converting data into tokens; the authors report this cut Transformer latency by 4.0–7.2 times and peak memory by 6.8–12.7 times across four EEG processing routes, without reducing accuracy.

On the SeizeIT2 seizure-detection test, the authors report that combining signals from separate encoders raised event sensitivity from 61.4% to 74.4% at 11 false alarms per hour. The paper lists code and model weights, but the supplied record does not include their links.

## The facts

- The EEG model has 30 million parameters and was pretrained on more than 15 million windows, according to the authors.
- The arXiv submission history dates the first version to November 6, 2026.

## Why it matters

For researchers processing EEG recordings, the reported memory savings could make these models less demanding to run, though the results do not establish how they perform in clinical use.

## Sources & references

1. [PhysioWave-T: Token-Efficient Wavelet Foundation Models for Biosignals](https://arxiv.org/abs/2611.08969) – arXiv, 2026-11-09

Last updated: 2026-11-09