Voice-Controlled Device Wake-Word Detection Using AEC Statistics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wake-word false triggering during audio playback in voice-controlled devices causes confusion and incomplete audio playback, as the system mistakenly pauses or attenuates audio when it detects a wake-word present in the audio stream, rather than one spoken by the user.
Innovation Solution
The use of acoustic echo cancellation (AEC) statistics to differentiate between a user-spoken wake-word and a wake-word present in the audio playback, by monitoring Echo Return Loss Enhancement (ERLE) and other AEC metrics to set a wake-word accept/reject flag, allowing the system to ignore false triggers and maintain uninterrupted playback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the device continuously monitors for wake-words during audio playback, then it can respond to user commands, but it causes false triggering when the playback audio contains wake-word-like sounds
Solution Approach 1:
The patent introduces an intermediary mechanism (AEC statistics analysis) between the wake-word detection and the response activation. Instead of directly triggering on detected wake-words during playback, the system uses AEC statistics as a mediator to verify whether the detected wake-word is genuine (user-spoken) or spurious (from playback audio), thereby resolving the false trigger problem while maintaining reliable wake-word detection
Solution Approach 2:
The system implements feedback by continuously monitoring AEC statistics and using them to dynamically adjust wake-word detection behavior. The AEC statistics provide feedback about the current playback state, which is fed back to the wake-word detection module to determine whether to accept or reject detected wake-words, creating a closed-loop control system that prevents false triggers
2Ease of operation
If the system pauses or attenuates audio playback upon detecting a wake-word, then it can process user commands, but it interrupts continuous playback when the wake-word is not user-spoken
Solution Approach 1:
The system performs preliminary analysis of AEC statistics before pausing or attenuating audio playback. By evaluating AEC statistics in advance, the system can determine whether a detected wake-word is genuine, and only then proceed to pause playback for command processing. This preliminary action prevents unnecessary interruptions to continuous playback while maintaining command processing capability
3Difficulty of detecting and measuring
If the device uses traditional wake-word detection during playback, then it can detect wake-words, but it cannot distinguish between user-spoken and playback-containing wake-words
Solution Approach 1:
The patent changes the parameters used for wake-word detection by incorporating AEC statistics (such as ERLE values) into the detection process. Instead of relying solely on traditional wake-word matching, the system monitors changes in AEC parameters to determine the source of detected wake-words, thereby improving measurement precision in distinguishing between user-spoken and playback-containing wake-words
Data Source
AI summary
Devices and techniques are generally described for control of a voice-controlled device using acoustic echo cancellation statistics. A reference signal representing the audio stream may be sent to an acoustic echo cancellation (AEC) unit. A microphone may receive an input audio signal and send the input audio signal to the AEC unit. The AEC unit may attenuate at least a part of the input audio signal. AEC statistics related to the attenuation of at least the part of the input audio signal may be determined over a first period of time. A wake-word in the input audio signal may be detected during the first period of time. A determination may be made that the wake-word is part of the playback of the audio stream based at least in part on the AEC statistics.