Turn abandoned patents
into your own filing
into your own filing
This invention describes a system that improves voice call quality by intelligently detecting who is speaking during a conversation. It analyzes microphone audio, comparing it to learned patterns of both the remote speaker and a general speaker to determine if one person is talking or if both are talking simultaneously. Based on this, the system adjusts how aggressively it removes echoes, prioritizing clarity of the local speaker's voice. The claims detail a method where the system learns to recognize the remote speaker from their playback audio and the local speaker from the microphone audio.
Why it matters: Filed in 2018, this invention predates significant advancements in deep learning for real-time audio processing and speaker diarization on consumer devices. Modern machine learning frameworks and increased edge computing power now make training and deploying sophisticated speaker models for dynamic echo cancellation more efficient and effective.
AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.
Reinvent this with AI
