DSP Group’s DBMD7 is a family of multi-core AI and DSP processors targeting audio/voice smart applications for a far-field voice user interface, voice communications and audio post processing. The DBMD7-2 supports two microphones, the DBMD7-4 supports four microphones, and the DBMD7-8 supports eight microphones. All feature superior far-field beam-forming and acoustic echo cancellation (AEC) to allow barge-in capabilities. These preprocessing front-end algorithms are optimized to maintain high performance of both the local/embedded and remote/cloud automated speech recognition (ASR) engines under tough conditions. Both single and multiple embedded ASR engines are supported to enable wake word (WW) detection, seamless query recognition, optionally local voice commands, and natural language processing (NLP). Together with full-duplex voice calls, the DBMD7 family meets all voice requirements and provides comprehensive solutions, including a complete software pack, debug, tuning tools and documentation.

The DBMD7 family supports a neural network framework that can be used to enhance sound detection. Various innovative power schemes support overall power efficiency.