Myosphere's NeuralBand 4 Debuts Sub-Millisecond Gesture Recognition, Promises to Replace Mouse and Touchscreen in Professional Workflows

Wearable neural interface startup Myosphere launched the fourth generation of its wristband product on October 20th. The NeuralBand 4 introduces a custom-designed neuromorphic processing chip that performs on-device gesture inference in under 0.4 milliseconds, moving the entire recognition pipeline offline and eliminating the cloud round-trip that constrained earlier generations. The company claims a 99.4% accuracy rate across 240 recognized hand and finger movements, including fine-grain gestures such as tracing individual letters in the air.

The Technology: Surface Electromyography Meets Neuromorphic Computing

NeuralBand 4 uses an array of 64 high-density surface EMG (sEMG) sensors arranged around the wrist — double the sensor count of its predecessor — to capture the electrical signals generated by motor neurons in the forearm muscles. These signals are processed by Myosphere's MX-4 chip, a neuromorphic processor that mimics the event-driven signaling of biological neurons. Unlike conventional chips that process data in fixed clock cycles, the MX-4 only activates when an sEMG signal change is detected, consuming an average of 18 mW during continuous use — comparable to a typical fitness band.

Professional Use Cases: Surgery, CAD, and Live Production

Myosphere's primary target markets are professional environments where fine motor control and sterility limit existing input methods. In surgical settings, the company has partnered with Singapore General Hospital to trial NeuralBand 4 for hands-free control of imaging systems during procedures. In computer-aided design, the company demonstrated real-time 3D model manipulation at Autodesk University last month, navigating and rotating a 20-million-polygon mesh without touching a keyboard or mouse.

For live broadcast production, NeuralBand 4 has been piloted by two major European sports networks to allow commentators to trigger instant replay clips and call up player statistics through discrete finger gestures, eliminating the need to look away from the action.

Privacy and Security Architecture

All gesture inference runs locally on the MX-4 chip; no sEMG data leaves the device. Myosphere has implemented a hardware-level encryption layer that isolates biometric data from the Bluetooth transmission stack. The company underwent a third-party security audit by Trail of Bits, whose report found no pathway for raw EMG data exfiltration under normal operating conditions. However, the audit noted that the BLE firmware interface presents a "moderate risk" attack surface for replay attacks targeting the gesture classification pipeline — a concern Myosphere says it will address in a December firmware patch.

Pricing and Availability

NeuralBand 4 ships in November 2027 in two configurations: the standard MX-4 at USD 349 and a Pro variant with an extended 96-sensor array for medical and research applications at USD 799. Both include the Myosphere SDK, supporting integration with Windows, macOS, and Linux via a unified HID-compatible driver that applications can use without custom API calls. Pre-orders opened October 20th, with first shipments targeting creative professionals and developer partners.