Ultra-Low Power Neuromorphic Front-End Chip for Voice Operation in Two-Way Radio Headsets
NeuroComm is a neuromorphic application-specific chip for mass-market headsets that turns any analog or digital two-way radio into a hands-free, intelligent, secure communication device.
Based on NASP, it addresses the problems of unreliable VOX and poor clarity without requiring expensive radio replacements.
Legacy two-way radios have challenges:
1Legacy VOX is error-prone, triggered by noise
2No speaker authentication may cause accidental activation or spoofing
3Manual group selection
4Poor clarity in noisy environments or EMI
1AI-powered voice detection activates transmission only by human voice
2Speaker recognition activates the radios only by known voices
3Smart Voice Control enables command recognition and hands-free group management
4Two-way voice extraction improves speech intelligibility
In the era where voice assistants are widely used in domestic settings, responding seamlessly to casual commands, AI-powered voice control applied to critical communication headsets can offer key advantages, such as reliable hands-free operation and a clear voice in noisy environments.
The key question is how to integrate AI processing into the headset. Cloud-based AI models cannot provide a reliable solution since cloud connectivity introduces excessive delays, and network unavailability could hinder critical communications.
Embedding a digital model into the headset requires a powerful CPU/NPU, bringing up power constraints and cost barriers.
A new type of embedded solution with ultra-low power consumption and latency is required for the new generation of smart headsets.
The NASP NeuroVoice chip enables the integration of multiple AI cores into a smart headset for radio communications. All this is possible within the power envelope of standard
headsets, consuming less than 200 µW. NASP neural cores can transform any headset into a smart, tactical-grade voice interface.
