-
-
LOGO
Inspiration
In today’s world, staying focused is becoming increasingly difficult. Notifications, messages, screens, tasks, and endless content constantly compete for our attention. We realized that focus should not depend on willpower alone. The environment around us should also understand and support our mental state.
That idea inspired Project:NOOS.
NOOS is designed around a simple question: What if our surroundings could understand our current cognitive state and adapt to help us reach the state we want?
Inspired by space travel, NOOS transforms complex brainwave data into an intuitive journey. Each planet represents a target mental state—such as activation, deep focus, or recovery—allowing users to choose where they want their mind to go next.
What it does
Project:NOOS is an AI-powered neuroadaptive environment that connects EEG sensing, personalized music generation, and smart lighting.
First, NOOS collects the user’s EEG signals in real time through a Muse headband. The system interprets these signals as understandable state dimensions, including focus, fatigue, stress, arousal, and emotional balance.
The user then selects a planet representing their desired mental state. For example, Mars represents energy and action, Neptune represents deep immersion, and Pluto represents recovery and reset.
NOOS compares the user’s current state with the selected target state and creates a personalized intervention plan. Based on that plan, it:
- Generates instrumental music adapted to the user’s state
- Adjusts the color, brightness, temperature, and movement of WiZ smart lights
- Divides the experience into phases such as stabilization, activation, deepening, and recovery
- Presents the entire process as an immersive space journey
NOOS does not diagnose the user. Instead, it translates the gap between the user’s current and desired states into changes in music and lighting.
How we built it
We built NOOS as a local multimodal system consisting of three main layers.
The frontend is a React and Vite application. It handles Muse connection through Web Bluetooth, EEG visualization, questionnaires, planet selection, music playback, lighting previews, and the space-travel experience.
The backend is built with Spring Boot. It connects the frontend to the Python AI engine, generated audio, ACE-Step music generation, and WiZ smart lights. It also manages API requests, local processes, audio delivery, and UDP-based hardware control.
The AI engine is implemented as an independent Python package. It:
- Parses raw EEG readings or frequency-band summaries
- Produces a structured recognition profile
- Maps each planet to a target-state vector
- Calculates the difference between the current and target states
- Selects an appropriate transition strategy
- Creates controlled music and lighting specifications
- Sends the music specification to ACE-Step for generation
Instead of asking the music model to generate freely, NOOS first creates a controlled specification containing BPM, tonal direction, rhythmic density, spectral brightness, instruments, texture, and elements to avoid. ACE-Step acts as the composer, while NOOS acts as the director.
Challenges we ran into
One of our biggest challenges was connecting technologies with very different runtime requirements. The complete experience crosses Web Bluetooth, a React frontend, a Spring Boot server, a Python AI engine, a remote GPU music-generation worker, and local smart-light hardware.
EEG signals also contain noise and uncertainty. We needed to avoid presenting unstable measurements as definitive conclusions, so we designed structured outputs that include confidence, signal quality, and limitations.
Another challenge was controlling generative music. Direct text-to-music generation can produce inconsistent results, so we created an intermediate specification layer that translates cognitive-state changes into bounded musical parameters.
We also had to keep music and lighting synchronized as one coherent intervention rather than treating them as separate effects. This required a shared phase-based plan for stabilization, activation, focus, and recovery.
Accomplishments that we're proud of
We are proud that NOOS goes beyond simply visualizing EEG data. It turns brain signals into an interactive, multimodal experience that changes the user’s environment.
We successfully connected:
- Real-time Muse EEG input
- AI-based state recognition
- Planet-based target-state selection
- Personalized intervention planning
- ACE-Step music generation
- WiZ smart-light control
- An immersive React-based space journey
We also built the AI engine so that its core recognition and intervention logic can run and be tested independently from the frontend and backend.
Most importantly, we created an understandable interaction model for complex neurotechnology. Users do not need to interpret raw brainwave graphs—they simply choose the state they want to move toward.
What we learned
We learned that building a neuroadaptive product is not only an AI problem. It requires careful coordination between signal processing, interaction design, generative media, backend orchestration, and physical hardware.
We also learned that uncertainty must be treated as part of the product experience. EEG-derived states should be communicated as estimates rather than medical facts.
Another important lesson was that generative AI becomes more useful when it operates within a well-designed control system. By defining state vectors, transition modes, musical constraints, and lighting parameters before generation, we achieved more consistent and explainable results.
Finally, we learned that metaphor can make advanced technology accessible. The space-travel concept allowed us to replace difficult numerical controls with an intuitive question: “Where do you want your mind to go next?”
What's next for Project:NOOS
Our next step is to turn NOOS into a closed-loop adaptive system.
During a session, NOOS will continuously observe EEG changes and adjust the environment in real time. If stress increases, the system could reduce brightness and musical harshness. If fatigue rises during a focus session, it could introduce subtle activation. If the user becomes overstimulated, it could slow the lighting motion and soften the sound.
We also plan to:
- Improve EEG calibration and artifact handling
- Personalize state models using long-term user data
- Expand the library of planets and target states
- Improve transitions between generated music segments
- Validate the experience through structured user studies
- Support more lighting and wearable devices
- Strengthen privacy-preserving, on-device processing
Our long-term vision is for NOOS to become an environment that does not merely demand attention, but actively helps people protect and recover it.
Log in or sign up for Devpost to join the conversation.