Two Electrodes an Inch Apart Scoring Sleep and Catching Micro Arousals From One EEG Channel

This application covers a physiological data acquisition assembly: a small housing holding a cluster of EEG electrodes placed close together, plus the processing that turns their output into a map of sleep stages and arousal states. The claimed device subtracts one electrode’s signal from another, builds a spectrogram, normalises it, sorts its segments into groups, and then works out which frequency features distinguish one group from another. It was filed by Neurovigil, Inc., the San Diego neurotechnology company founded by Philip Low, and it is the fourth Australian divisional in a chain reaching back to a US provisional application from 2013.

The Problem

Conventional EEG is a laboratory procedure. The specification describes the usual arrangement plainly: dozens of electrodes placed all around a person’s head, positioned precisely on scalp locations in an effort to improve the ratio of signal to noise. That setup takes time, needs a trained operator, and confines recordings largely to clinical settings, and even then, the specification notes, EEG is still widely criticised for poor spatial resolution and a low signal to noise ratio.

There is a specific electrode problem underneath the general one. EEG works by differentially amplifying the signals from an active electrode and a reference electrode, so the reference is traditionally placed far from the active site, somewhere electrically quiet. Putting the two close together was avoided on the grounds that it induces distortion, and that a reference electrode sitting over active cortex would itself record neural activity and cancel out the high frequency signals of interest. That convention is what forces the large head footprint.

Sleep scoring itself is coarse. Sleep is conventionally divided into stages 1 to 4 and a rapid eye movement stage, and awakenings are treated as a binary transition from asleep to awake. The specification argues that brief arousals, including micro arousals, go unrecognised entirely as a result, even though they are what degrades sleep quality.

What This Invention Does

Claim 1 defines a physiological data acquisition assembly: a housing carrying a cluster of electrodes comprising an active electrode and a reference electrode in close proximity to each other, with a processor and electronic circuitry that performs a specific sequence. It receives the reference signal, generates a differential signal by subtracting the reference from the active electrode to reduce noise, and generates a spectrogram from that differential signal.

The rest of the claim is the analysis. The spectrogram is normalised one or more times, and each of multiple segments of the normalised result is assigned to a group drawn from a set of groups. For each group the circuitry determines one or more frequency spectrum features, either segment by segment or collectively across the group. It then identifies one or more frequency signatures, defined in the claim as the frequency spectrum features that distinguish one group from the others. The claim closes by specifying what the groups must be: a set of arousal states, a set of sleep stages, or a set made up of both.

The physical scale sits in the body of the specification rather than in claim 1. A single small device houses the active electrode, the reference electrode and optionally a ground electrode, with a footprint under six inches and in some cases under four inches on a side, and no more than three inches between any pair of electrodes.

Two derived measures do the discriminating. A strong frequency is the frequency carrying the highest normalised power in a time bin, and its distribution shifts with sleep stage. A fragmentation value comes from the temporal or spectral gradient of the spectrogram and identifies the frequency showing the most modulation, which the specification associates with sleep stage disturbance. Because stages can be assigned within very short windows, arousals can be detected by counting transitions and variability inside a run of windows. The specification reports automated arousal detection agreeing with manual scoring at 98.4 per cent in one worked example.

Key Features

• Clustered active and reference electrodes. Both electrodes sit in one housing within inches of each other, reversing the conventional practice of placing the reference far away on quiet scalp.
• Repeated spectrogram normalisation. A z score is taken for each frequency across all time bins, then for each time bin across all frequencies, alternating until a normalisation factor falls below a threshold. This is what recovers usable structure from a recording that would otherwise be dismissed as noise.
• Group specific frequency signatures. Rather than matching against fixed templates, the assembly derives the frequency spectrum features that distinguish its own groups from one another.
• Short window sleep staging. Assigning stages within very short time blocks is what allows brief arousals to be detected at all, instead of only full transitions to wakefulness.
• A second use for the same signal. The specification extends the same processing to electromyography recordings, mapping muscle signal features onto cursor movement in an on screen speech assistance interface so a person can select letters and words without using their hands.

Who Is Behind It

The applicant is Neurovigil, Inc., a neurotechnology company based in La Jolla, San Diego. The sole inventor is Philip Low, the company’s founder, chairman and chief executive, who started it in 2007 after a doctorate in computational neurobiology at the University of California, San Diego for work done at the Salk Institute for Biological Studies. His thesis work produced the sleep scoring algorithm the company’s iBrain single electrode recorder is built around, and the device has been used in pharmaceutical trials to record sleep EEG before and after subjects take medication.

Low is also publicly known for a 2012 research collaboration with Stephen Hawking, who had motor neurone disease. Hawking supplied EEG recordings made with the iBrain, and Low presented an analysis at the Francis Crick Memorial Conference in Cambridge in July 2012, checking the recordings for repeatable patterns when Hawking was asked to imagine moving his hand. It is worth being precise about what that was: a scientific demonstration and an attempt to establish a signal, reported at the time as early stage work, not a communication product that went into use.

The chain behind this filing is long. AU 2026202122 is a divisional of Australian application 2023241380 filed 6 October 2023, itself a divisional of 2021250913, itself a divisional of 2019204112, itself a divisional of 2014337565 filed 14 October 2014, which was the Australian national phase of PCT/US2014/060489. That international application claims priority from US provisional application 61/890,859, filed 14 October 2013, which makes the United States the priority country.

Why It Matters

The gap this specification aims at has widened rather than closed since 2013. Consumer sleep tracking has grown into a large category built mostly on movement and heart rate rather than brain activity, which is cheap and comfortable but cannot distinguish sleep stages with confidence. Clinical polysomnography remains the reference standard and remains a night in a laboratory wired to a rack of channels. A device producing stage level output from two electrodes in one housing sits between those two.

Micro arousals are the part most likely to matter clinically. Fragmented sleep with frequent brief arousals is a recognised feature of conditions including sleep apnea, and counting those events by hand across a night is slow work that scorers do not always agree on. Automating detection at a finer resolution than manual scoring uses changes what can be measured, though the accuracy figures here are self reported and drawn from a small number of nights.

The cursor control branch belongs to a field that has moved a long way. Implanted brain computer interfaces have since demonstrated far higher communication rates for people with amyotrophic lateral sclerosis and similar conditions, but they require surgery. A surface electrode assembly reading muscle signals is a much smaller intervention at a different point on the same trade off. Filing a fourth Australian divisional twelve years after the priority date is what a company does to keep claim scope alive while the underlying technology is still finding its commercial shape.

Related Concepts

• Polysomnography – the laboratory sleep study this device is positioned as an alternative to.
• Sleep apnea – the condition whose repeated brief arousals this arousal detection is most obviously aimed at.
• Non rapid eye movement sleep – the staged structure that the claimed groups are drawn from.
• Signal to noise ratio – the measure the conventional many electrode arrangement exists to improve.
• Tracheotomy – the procedure whose overnight risks the specification names as one target for arousal monitoring.

---

AU 2026202122 was published in the Australian Official Journal of Patents on 9 April 2026 and is open for public inspection. Patent applications represent inventions that are sought to be protected and do not necessarily reflect commercially available products.