AON1010™ belongs to the highly optimized AONVoice™ Neural Network cores for Voice and Audio recognition. This solution is optimized for processing microphone data for applications including voice and sound recognition.
The solution consists of:
* AON1010™ AI processing engine delivered in Verilog RTL. The RTL IP is suitable for logic synthesis targeting ASIC products as well as FPGA.
* AON1010™ AI software.
AI processing engine
Overview
Key Features
- Voice control, Context detection and Speaker ID
- Always-on Multi-Wake-word detection
- On-device Voice command recognition (‘Music play’, ‘Music stop’, etc.)
- Ultra-Low power for always-listening voice-enabled applications
- Adaptive voice activity detection
- Acoustic Event Detection (AED)
- Context-awareness
- Speaker Identification
- High Accuracy in Noisy Conditions
- High accuracy in background noise and reverb.
- Speaker independent
- Robust to variabilities within user, between-user & in environment
- Scalable, detects multiple commands concurrently
Video
AON1100 Text-Dependent Speaker Identification in Noisy Environment
Technical Specifications
Related IPs
- AI processing engine for Wake Word, Voice Commands, Acoustic Event Detection, Speaker ID and Sensors
- AI processing engine
- High-performance 32-bit multi-core processor with AI acceleration engine
- High-performance 64-bit RISC-V architecture multi-core processor with AI vector acceleration engine
- Small-size ISP (Image Signal Processing) IP ideal for AI camera systems.
- Enhanced Neural Processing Unit for safety providing 32,768 MACs/cycle of performance for AI applications