Bose holds a granted patent covering exactly this feature: on-device neural network environment classification driving ANC parameter adjustment. This claim is blocked as-written.

Sony's patent covers a sub-3 ms feedback loop using analog prediction. Your "dual-layer" framing is not in their claims, which creates design room, but the latency threshold overlaps. You may be able to distinguish on the dual-layer architecture.

Apple's beamforming patent covers on-device isolation via beamforming arrays, which is mechanistically different from your approach (speaker embeddings without beamforming). No granted claim appears to block on-device embedding-based isolation using a single microphone capsule.

Methods and systems for active noise cancellation that adjust cancellation parameters based on detected environmental context using a neural network classifier trained on ambient audio.

System for isolating individual speaker audio from a multi-speaker input using adaptive beamforming and speaker embedding models running on embedded DSP hardware.

Feedback-based active noise cancellation architecture achieving sub-5 ms end-to-end latency through dedicated analog front-end and prediction-based filter coefficient update.