When a user speaks to a smart speaker or voice assistant while the device is playing back a Text-to-Speech (TTS) response, the microphone captures an overlapping mixture of the user's speech and the reverberant TTS playback echo. Textual Echo Cancellation (TEC) cancels the TTS echo using the source text of the TTS playback (< 0.1 KB side input) and reconstructs the clean user speech for Automatic Speech Recognition (ASR).
examples/sample_1..4_{tec_single,tec_multi,aec_single}.wav) for each selected model. Custom uploads or modified text in this static browser page run an in-browser WebAudio text-conditioned spectral canceller; run python3 app.py or scripts/inference.py locally for full Lingvo/TensorFlow checkpoint inference on arbitrary files.
| Approach | Audio Input to ASR | Side Input Payload | ASR Recognition Result |
|---|---|---|---|
| Run the demo to compare ASR results before and after Textual Echo Cancellation. | |||
Real 0 dB SNR reverberant mixtures (RT60 = 0.25 s) from the LibriTTS test-clean + LJ Speech evaluation set. Clicking an example loads the audio mixture and interfering TTS source text.
| Example | Interfering TTS Playback Text (Side Input to Cancel) | Ground-Truth Clean User Query (Target) | Action |
|---|---|---|---|
| Sample 1 | To obtain possession of this, Thurtell with his two associates resolved to kill him. |
"\"We've tortured each other enough for tonight." | Load & Run → |
| Sample 2 | As he told the driver, quote, Let's get out of here; we are hit, end quote, |
"The chair was empty; but he knew." | Load & Run → |
| Sample 3 | Carro reported that when questioned about his mother Lee said, quote, |
"Have we followed that road?" | Load & Run → |
| Sample 4 | The Commission has concluded that on April ten, nineteen sixty-three, |
"The parliament and the Scots laid their proposals before the king." | Load & Run → |