本demo主要基于深度学习的aec回声消除开发，用神经网络更精确区分回声与近端语音，减少误判。在线性滤波后用神经网络做精细化的残余回声抑制。从效果和性能全面超越webrtc aec3.

资源链接：windows平台音频ai-aec高级回声消除demo资源-CSDN下载

用法：
AEC4 Demo - AEC4-API v1.0.0 (Advanced AEC)
AEC4 Demo v1.0 - AEC4-API v1.0.0 (Advanced AEC)

Usage:
  aec4.exe file   <far.wav> <near.wav> <out.wav> [rate] [nlp]
  aec4.exe stream <far.wav> <near.wav> <out.wav> [rate] [nlp]
  aec4.exe synth  <out.wav> [duration] [rate] [nlp]
  aec4.exe bench  [duration] [rate] [nlp]

Modes:
  file   - Offline file processing (pre-buffer render)
  stream - Real-time stream simulation (interleaved render/capture)
  synth  - Generate synthetic test signals
  bench  - Performance benchmark

Examples:
  aec4.exe file far.wav near.wav out.wav 16000 4
  aec4.exe stream far.wav near.wav out.wav 16000 4
  aec4.exe synth test.wav 10 16000 4
  aec4.exe bench 60 16000 4

自带测试case: near_end.wav 和 far_end.wav  

笔记本环境测试， CPU: AMD Ryzen 5 7640HS w/ Radeon 760M Graphics (4.30 GHz)