Speaker Output Attenuation During Barge-In Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems face challenges in accurately recognizing user input when the user interrupts audio output from a text-to-speech system, media player, or navigation system, as the microphone captures both signals, leading to inaccurate recognition due to feedback.
Innovation Solution
A computer-implemented method and system that receive a signal containing both the output of a speaker device and an additional audio signal from a user. By using a model trained to identify the speaker device output, the system determines if the additional audio signal corresponds to a user utterance and reduces the audio output level of the speaker device accordingly, allowing for accurate recognition of the user's input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the speaker device continues audio output during user speech, then the audio output is continuous and uninterrupted, but the speech recognition accuracy deteriorates due to feedback from the speaker output
Solution Approach 1:
The system performs preliminary detection of user speech intent before the speaker output interferes with recognition. The dialog manager monitors the audio signal continuously and proactively identifies when a user intends to speak, initiating audio suppression in advance to prevent feedback contamination of the recognition signal.
Solution Approach 2:
The harmful speaker audio output is extracted and separated from the microphone input signal. The system isolates the user speech component from the combined audio signal by suppressing the speaker output path when user speech is detected, effectively removing the harmful feedback element while preserving the user's spoken input.
2Measurement precision
If the audio output level is reduced during user utterance, then the speech recognition accuracy improves, but the audio output continuity is interrupted
Solution Approach 1:
The audio output level is dynamically adjusted based on real-time detection of user speech. The system transitions the speaker device between different operational states (full output, reduced output, or muted) depending on whether user speech is detected, allowing flexible adaptation to maintain both audio continuity and recognition accuracy.
Solution Approach 2:
The audio output parameter (volume level) is changed conditionally based on speech detection. When user speech is detected, the system modifies the audio output level parameter to a reduced state, and restores it when speech ends, thereby adjusting the parameter to balance recognition precision with audio output continuity.
3Measurement precision
If the system suppresses speaker output to improve recognition, then recognition accuracy improves, but the dialog efficiency may worsen due to additional processing steps
Solution Approach 1:
The speech detection and audio suppression functions are merged into the existing dialog manager component. Rather than adding separate processing modules, the system integrates the intent detection and audio level control within the existing dialog management architecture, reducing overhead and maintaining dialog efficiency while improving recognition accuracy.
Data Source
AI summary
The technology described in this document can be embodied in a computer-implemented method that includes receiving, at a processing system, a first signal including an output of a speaker device and an additional audio signal. The method also includes determining, by the processing system, based at least in part on a model trained to identify the output of the speaker device, that the additional audio signal corresponds to an utterance of a user. The method further includes initiating a reduction in an audio output level of the speaker device based on determining that the additional audio signal corresponds to the utterance of the user.