As a base of our analysis, we chose frontal alpha rhythm sensor readings for a baseline metric to assess the current state of a person’s mind during meditation, because the activity of alpha rhythms are proven to be associated with a state of relaxation and calmness, which characterizes the desired meditation effect.

Quantization of alpha levels into definite gradations associated with brain states is a very hard task due to a high degree of individuality in brain activity footprints. Therefore we decided to track consistency of changes in the overall alpha activity. When activity significantly increased over some period of time, indicating consistent meditation without distractions, we altered playback of thematic video, soundscape and LED-light sequences designed to indicate deeper meditation. Alternately, if there was a noticeable decrease in recent frontal alpha activity, meaning a person was likely unable to maintain concentration, we altered thematic playback to indicate shallower meditation. If no significant change in activity occurred, we did not alter thematic playback.

Our EEG band (InteraXon MUSE) measures activity levels 10 times per second. Our algorithm stores around 30 seconds of latest data at any given moment, and every 5 minutes calculates the trend of the activity for the next 30 seconds with the ARIMA model for time-series forecasting. Then logic is applied to switch between sound and visual themes according to the result of this calculation.

EEG (electroencephalography) is a method to record electrical activity in a brain. With this non-invasive technique we can easily observe a person’s electrical activity using a Muse devise, a headband with four electrodes which monitors a brain's activity in frontal and temporal lobe (forehead and behind ears area respectively). The electrical activity is divided into bands by frequency, known as alpha, beta, gamma, delta and theta waves.

This chart depicts alpha, beta, gamma, delta and theta waves in a frontal lobe, recorded during 15 minutes of a meditation (quivering lines). Smooth solid lines show an average of the corresponding waves. The same design is used in following charts.

Alpha wave activity is associated with meditative states, so we are mostly interested in alpha.

 

This chart depicts alpha waves in each of four electrodes: left ear, right ear, left frontal and right frontal. Two frontal electrodes show pretty similar patterns. Alpha activity in a frontal lobe best reflects meditative states, so in our algorithm we used averaged alphas from two frontal electrodes (coral and pink in the chart).

 

This is the same data as used in a previous plot, but only the first 60 seconds. Here you can see the waves in detail.

 

This is a chart with records of 8 experienced meditators. Alpha wave is highly affected by different factors, as age, time of the day, eyes closing, weariness, etc., but nonetheless it is possible to discover common patterns.

 

This chart illustrates the work of our algorithm. Every three minutes it analyzes the previous alpha waves records and decides whether the meditation deepens, lightens or stays at the same level. Color depicts these stages. Every meditator starts at a stage 1, most of the meditators do not exceed the 3rd stage, but we have seen very deep meditative states, level 5, the highest one. Solid black line shows an average of frontal alpha, which was used in the algorithm.

In this plot meditator started at the stage 1 (green section), then, after three minutes algorithm decided that his meditation went deeper and switched to the stage 2 (coral section). The same happened in the next two 3 minutes intervals, violet and pink sections. You can see that alpha was slowly growing over this period of time. But after 12 minutes alpha became lower, the algorithm detected this and switched from 4th stage (pink) back to 3d (violet).