At Embedded World 2024, My Voice AI showcased its cutting-edge speaker verification technology powered by the Syntiant NDP120 chip. This system not only recognises what is said but also identifies who is speaking, offering both security and personalisation. By using a neural network to match an enrolled voiceprint to the speaker, it ensures only verified users can access specific features or content.
My Voice AI, a UK-based startup, has developed one of the world’s smallest speaker verification engines, with their unique voiceprint being less than 1KB. The technology operates entirely on the edge, eliminating the need for data transfer to external servers for verification. This focus on on-device processing enhances both security and privacy, as users’ voice data remains secure.
How My Voice AI’s Technology Works
The technology revolves around enrolling users directly on the device. When a user first interacts with the system, the engine captures and processes unique vocal characteristics to create a voiceprint. This voiceprint is then utilised for future verification. In normal use, the system compares the speaker’s live voice to the pre-enrolled voiceprint. By utilising text-independent verification, the speaker is recognised based on their unique vocal patterns, irrespective of the specific words spoken. This versatility ensures seamless security and access control across various devices without requiring predefined voice commands.
Key Applications: Security and Personalisation
My Voice AI’s technology offers a wealth of possibilities across different markets. One notable application is access control. Devices like smart locks or hearable devices (such as earbuds) can be locked and unlocked using only the user’s voice. In the case of earbuds, which often lack traditional user interfaces, the voice becomes the primary interface for managing access. This allows users to lock devices and ensure only authorised individuals can use them.
Another major application is personalised content delivery. By identifying the speaker, the system can deliver tailored media content. In a smart home environment, when different family members request media on a smart TV, the system can distinguish between users and provide personalised content based on their preferences. This level of user-specific personalisation goes beyond typical household-level customisation. A family member can, for example, open their YouTube or Netflix account simply by issuing a voice command to the TV remote.
Scalability and Efficiency
My Voice AI’s engine can accommodate multiple users on a single device, with up to 20 users tested so far. The engine’s small footprint—occupying only 500KB in total, with a model size of 300,000 parameters—makes it highly scalable across a wide range of applications. Whether in a large family or a shared office environment, the system seamlessly handles numerous users without sacrificing performance.
Real-Time Demonstration: Syntiant NDP120 Chip Integration
At the event, My Voice AI demonstrated its speaker verification technology on the Nicola voice module, powered by the Syntiant NDP120 processor. This hardware demo showcased how the neural network processes live speech, verifying the user in real time. The system displayed a green light for authenticated access and a red light for denied access, demonstrating its real-world potential in various contexts, such as smart devices and personalised entertainment systems.
The demo validated the technology’s practicality and accuracy, illustrating how My Voice AI’s system can seamlessly integrate with IoT devices to make homes, offices, and personal electronics smarter, more secure, and more user-friendly.
The Future of Voice-Activated Systems
My Voice AI’s speaker verification technology leads the way for the next generation of voice interfaces. By enabling secure and personalised experiences across a wide range of devices—from smart home systems to wearables—the technology is redefining how people interact with their devices. With voice becoming an increasingly essential interface in many applications, My Voice AI’s focus on on-device processing and personalisation sets it apart, ensuring that users enjoy both convenience and privacy in their interactions with technology.
As this technology continues to evolve, its potential to reshape the way we engage with devices is significant, paving the way for more intuitive, secure, and personalised tech experiences.