Apple develops approaches for AirPods with EEG sensors

This article may contain affiliate links. Learn more on our Partner Program page.

Make Apfelpatient a preferred source One click – and you'll see us more often on Google

Apple continues its efforts to transform its devices into health tools. A new study by the company shows how an AI model can learn the structure of electrical activity in the brain from unlabeled EEG data. This is interesting because Apple is simultaneously working on hardware that could measure such signals directly in the ear. Together, these developments paint a picture of how AirPods could one day not only play music, but also recognize sleep phases, detect brain abnormalities, or analyze neurological activity.

Current research focuses on whether AI models can teach themselves to recognize temporal patterns in EEG signals. This would be advantageous because current models rely heavily on human-annotated data. These annotations are expensive and time-consuming, but necessary for applications such as sleep stage analysis or seizure detection. Apple's approach explores whether this dependency can be reduced. In parallel, a 2023 patent describes how AirPods could be equipped with multiple biosensors. Officially, these two developments are unrelated. However, they demonstrate that Apple is pursuing research on two fronts simultaneously, which could complement each other effectively.

PARS as a new approach for EEG signals

The study, titled "Learning the relative composition of EEG signals using pairwise relative shift pretraining," introduces PARS, which stands for Pairwise Relative Shift. This approach is a form of self-supervised learning. The model is not designed to reconstruct missing signals, but rather to predict the temporal separation between two randomly selected EEG segments. This allows the AI to learn longer patterns in neural signals, not just local structures.

The study explains that current EEG SSL methods mostly rely on masked reconstruction. Masked autoencoders fill gaps in the signal and focus on short time intervals. Position prediction has been scarcely used so far, even though it is better suited for long-term dependencies. PARS changes this by forcing the encoder to understand the relative temporal composition of the signals.

The researchers tested PARS across several EEG benchmarks. The pre-trained model outperformed or matched existing methods in three out of four tasks. These included sleep phase detection, abnormal EEG pattern recognition, seizure detection, and motor imagery tasks. The results suggest that PARS could be a novel form of self-supervised EEG learning that requires fewer labels while still delivering strong performance.

The role of ear EEG in the study

The training program used, among other data, the EESM17 dataset (Ear-EEG Sleep Monitoring 2017). This dataset contains nighttime recordings from nine participants using a portable 12-channel ear EEG system and a 6-channel scalp EEG system. Ear EEG uses different electrode placements than traditional scalp systems. Nevertheless, many clinically relevant signals can be recorded, including sleep stages and specific seizure patterns.

This makes ear-based EEG interesting for wearables. If brain signals can be reliably measured at the ear, devices like AirPods offer an ideal basis for discreet biometric sensors. While the study itself doesn't mention AirPods, it shows that ear-based data is of sufficient quality to train a self-monitoring model like PARS.

The 2023 patent and Apple's hardware vision

In 2023, Apple filed a patent describing a wearable electronic device for measuring biosignals. The patent explicitly mentions that brain activity can also be measured in or around the outer ear. It cites advantages such as improved portability and reduced electrode visibility compared to scalp EEG. However, it also outlines the challenges. These include the need for individual customization to the ear to ensure reliable electrode contact. This is difficult because ear shapes vary considerably from person to person and can even change over time.

The patent describes a solution. Instead of using less precisely placed electrodes, Apple equips the device with more electrodes than necessary. An AI model evaluates these based on impedance, noise level, contact quality, and the distance between active and reference electrodes. It then assigns weights and constructs a single, optimized waveform from all the signals. The patent also describes tap or press gestures on the earbud to start or stop measurements, as well as various design options for sensor placement.

Apple mentions that the biosignals obtained could be used for applications such as sleep monitoring or seizure detection. These examples correspond almost exactly to the tasks that were also investigated in the PARS study.

Link between research and the possible future of AirPods

The study and the patent are not officially directly related. Nevertheless, a clear trend is emerging. Apple is researching methods to efficiently interpret EEG data and, at the same time, hardware that can record such signals at the ear. This is further supported by the current state of the AirPods. The AirPods Pro 3 already feature a PPG sensor for heart rate measurement. Apple is thus expanding the biometric capabilities of its earbuds step by step.

It's therefore not far-fetched to imagine that future AirPods could contain EEG sensors. This would transform them into wearables with a significantly expanded range of functions. In addition to music and communication, they could provide real-time neurological information or support important health functions.

Why AirPods could become more than just audio devices

The study shows that an AI model can learn the temporal relationship between segments of raw, unannotated EEG data. This increases efficiency and makes the analysis of brain data more accessible. At the same time, the 2023 patent reveals that Apple is working on suitable ear sensors that could capture EEG signals. Whether this will ultimately end up in a product like AirPods remains to be seen. However, it is clear that Apple is increasingly integrating AI, sensors, and wearables. This could eventually transform AirPods into devices that not only deliver audio but also provide deeper insights into one's own health. (Image: Shutterstock / Ivan_Shenets)