Understanding our approach to creating personalized calm for autistic children
Autistic children often experience uncontrollable emotions and heightened hyperactivity, which can disrupt their ability to relax and get adequate sleep. Lack of sleep further exacerbates behavioral and emotional challenges, creating a cycle that is difficult to break.
To address this issue, we've developed a machine learning (ML) model that analyzes the brain waves (EEG data) of autistic children. By extracting meaningful parameters from the EEG signals, the system generates personalized, calming content to help the child transition from a hyperactive state to a calm, restful one.
Generated dynamically based on EEG-derived parameters, with specific musical notes or chords (e.g., A-sharp, C-sharp) tailored to the child's current brain state.
Stock footage of serene environments such as rain, flowing water, or nature scenes designed to visually complement the music and create an overall calming experience.
We collect EEG signals from the child using a wearable EEG device and analyze the signals to infer key parameters related to brain activity, such as:
We calculate averages for each type of brainwave to determine the overall state:
We use the EEG parameters to determine the emotional and activity state of the child, then generate music based on their specific needs:
Range: 60–80 beats per minute (bpm). Higher arousal lowers the tempo to slow the music and counteract hyperactivity, while a relaxed profile keeps it closer to 80 bpm for gentle engagement.
The dominant wave reflects the child's baseline state, personalizing the key to their typical brain activity. Different keys evoke different emotional responses to match or counteract the child's state.
These adjust MIDI notes for each EEG interval based on specific wave strengths, ensuring dynamic adaptation:
Range: 40–80 (C2 to G5). Alpha and theta drive pitch upward as they indicate calm states; arousal waves lower pitch during hyperactivity, promoting relaxation with deeper tones.
Range: 1–5 semitones. High arousal reduces the step size for stability and simplicity, while low arousal allows larger, flowing intervals.
Range: 0.5–2.0 seconds. Higher relaxation waves lengthen notes, slowing the music to calm the child, while high arousal shortens durations.
Range: 40–80. Ties velocity to the child's peak relaxation capacity (max_alpha), personalizing loudness to their strongest calm state.
These refinements further personalize the music based on statistical extremes and tendencies:
Detects if peak arousal exceeds 60%, indicating significant hyperactivity, and lowers the pitch range to emphasize deeper, more grounding tones.
If the most frequent state is arousal, reverses step direction to create descending melodies that musically "wind down" arousal.
Complement the personalized music with appropriate visual content:
Every formula includes all five wave types (directly or via ratios), capturing their full influence rather than reducing them to a single index.
Global parameters reflect the child's overall EEG profile via averages and maxima, while per-interval parameters adapt to moment-to-moment changes.
Lower pitches, smaller steps, and longer durations during high arousal align with music therapy principles to boost alpha/theta and reduce beta/gamma.
Maxima adjust intensity, modes influence contour, and averages set the baseline, ensuring a rich, data-driven output.
Upload your EEG data to generate personalized, calming content for your child
Get Started