@Swapnil_The_Curious

The NeuroVox companion application is a custom-built Android and iOS app that serves as the complete output interface for the device. Classified muscle gesture data is received from the ESP32-S3 over Bluetooth Low Energy and passed directly to an integrated Claude Opus 4.6 API pipeline, which interprets the raw gesture intent and generates contextually appropriate, grammatically complete conversational responses on behalf of the user. The app displays these AI-generated responses as readable text in a clean minimal interface, supports artifact rendering for structured outputs, logs full conversation history within each session, and handles text-to-speech audio playback of the Claude-generated responses through the phone speaker so the user's communication is both visible and audible to everyone in the room without requiring any on-device audio hardware beyond the wearable speaker for alert purposes.

NeuroVox v2.0 is built on a custom 220×50mm 4-layer FR4 PCB in a long narrow strip form factor that wraps conformally around the forearm. The board runs four AD8232 EMG front-end ICs across four independent muscle channels, an ESP32-S3 module with native USB support, an ADS1299 8-channel biosignal ADC for high-resolution EMG acquisition, an INA333 instrumentation amplifier, a haptic feedback motor for gesture confirmation, a BNO055 IMU for motion artifact detection and rejection, dry reusable Ag/AgCl electrode contacts integrated directly into the PCB edge fingers eliminating separate lead wires entirely, a MCP73831 LiPo charger, a TPS61023 boost converter, a 5000mAh LiPo battery, a MAX98357A I2S amplifier driving a 3W 4-ohm speaker, RGB status LEDs, and a USB-C port for charging and firmware flashing.