SteadyPath

Inspiration

People living with Parkinson’s, like Will's Grandma, often struggle to monitor their condition consistently — from tremor intensity to medication adherence and symptom journaling. Most existing tools are either too manual, not tailored to their experience, or don’t integrate real-time input like speech. Typing and pressing small buttons can be very difficult and frustrating. We wanted to build a system that felt natural, non-invasive, and helpful, empowering users to track their condition with minimal friction.

What it does

Our app allows Parkinson’s patients to:

• Track tremor severity in real-time using accelerometer data and signal processing (frequency analysis on time domain signal)
• Use their voice to log symptoms and journal experiences — powered by AI speech-to-text
• Log medication usage, and link it directly to tremor events and journal entries
• Visualize symptom progression over time through clean, accessible charts

The experience is mostly voice-controlled, making it easier to use even during symptom flare-ups.

How we built it

• React Native with Expo for cross-platform mobile development
• SQLite (via expo-sqlite) for local session and medication logging
• DeviceMotion API for real-time tremor frequency and amplitude calculation
• Fast Fourier Transform (FFT) for signal analysis and tremor quantification (intensity and dominant frequency)
• OpenAI's Whisper (via AssemblyAI) for converting user speech into structured journal data
• Modular service architecture for handling database queries cleanly

Challenges we ran into

• Signal analysis: Accurately extracting tremor frequency and intensity from noisy accelerometer data was difficult.
• Speech integration: Balancing latency and accuracy when using real-time transcription while keeping the UI responsive.
• Voice-first UX: Ensuring the app is usable with minimal touch, which meant rethinking interactions, feedback, and session flow.

Accomplishments that we're proud of

• Built a real-time tremor analyzer using raw accelerometer signals and FFT
• Seamlessly integrated AI-powered journaling from speech input
• Created a voice-friendly, non-clinical user interface
• Fully linked symptom data, medication usage, and personal reflections into one timeline

What we learned

• How to perform signal analysis and frequency extraction with FFT on mobile devices using built in accelerometers.
• How to design and implement speech-first UX patterns
• How to use lightweight tools like sqlite and react-native to design functional mobile apps.

What's next for Parkinson’s App

• Adding alerts and reminders for medication based on trends and missed logs
• Training an ML model to correlate tremor patterns with medication effectiveness
• Integrating remote data sync so caregivers and clinicians can view insights
• Supporting multi-language voice input to make the app more accessible globally

Built With

• assembly-ai
• javascript
• react-native
• sqlite