Self-trigger Prevention in Voice Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Electronic devices face the challenge of self-triggering due to residual echo from output audio, which can incorrectly detect wakewords and trigger voice commands, leading to unwanted activation.
Innovation Solution
Implementing a system that performs double-talk detection and suppression of wakewords during active playback by generating correlation data between output and reference signals, allowing the device to differentiate between user speech and machine-generated audio, thereby preventing self-triggering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If echo cancellation is performed to remove residual echo from output audio, then the quality of audio output is improved, but self-triggering may still occur when residual echo exceeds energy limits
Solution Approach 1:
The system performs preliminary double-talk detection by analyzing correlation data between output audio and reference signals before wakeword detection occurs. This advance detection allows the system to identify machine-generated audio patterns and suppress wakeword detection during playback, preventing self-triggering before it can occur.
Solution Approach 2:
The system introduces an intermediary double-talk detection mechanism that acts as a gatekeeper between the audio output and wakeword detection. This intermediary layer analyzes correlation data and controls whether wakeword detection should proceed, effectively mediating between echo cancellation output and wakeword recognition to prevent false activations.
2Ease of operation
If the device performs wakeword detection on all audio input, then voice command responsiveness is improved, but false activation occurs due to machine-generated audio
Solution Approach 1:
The system applies different quality control to different portions of audio input by analyzing correlation data for each audio frame. When machine-generated audio is detected through double-talk detection, wakeword detection is selectively suppressed for those specific frames while remaining active for frames containing genuine user speech, creating locally differentiated detection behavior.
Solution Approach 2:
The wakeword detection system dynamically adjusts its sensitivity based on real-time correlation analysis. The system transitions between active detection and suppressed detection states depending on whether double-talk conditions are present, making the detection behavior adaptive and dynamic rather than static, thereby maintaining responsiveness while preventing false activation.
3Measurement precision
If correlation analysis is performed continuously to distinguish user speech from machine-generated audio, then self-trigger prevention accuracy is improved, but computational complexity increases
Solution Approach 1:
The system performs correlation analysis at a reduced rate and with partial data rather than continuous full-analysis. By computing correlation data selectively for audio frames during active playback and using simplified comparison logic against reference signals, the system achieves sufficient differentiation accuracy without the full computational burden of continuous speech analysis.
Data Source
AI summary
A system configured to perform self-trigger prevention to avoid a device waking itself up when a wakeword is output by the device's own output audio. For example, during active playback the device may perform double-talk detection and suppress wakewords or other device-directed utterances when near-end speech is not present. To detect whether near-end speech is present, an Audio Front End (AFE) of the device may perform echo cancellation and generate correlation data indicating an amount of correlation between an output of the echo canceller and an estimated reference signal. When the correlation is high in certain frequency ranges, near-end speech is not present and the device may suppress the utterance. When the correlation is low, indicating that near-end speech could be present, the device does not suppress the utterance and sends the utterance to a remote system for speech processing.