Speaker Audio Suppression for Barge-In Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems face challenges in accurately recognizing user input when it overlaps with audio output from text-to-speech systems, media players, or navigation systems, due to feedback issues that distort the captured audio and affect recognition accuracy.
Innovation Solution
A computer-implemented method and system that uses a model trained to identify the output of a speaker device to detect user utterances, allowing for a reduction in the audio output level of the speaker device, thereby suppressing the background audio and enhancing the recognition of user input, using techniques such as i-vector based models and neural networks to differentiate between user voice and device output.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the speaker device continues to output audio at normal levels, then the audio output quality is maintained, but the speech recognition accuracy deteriorates due to feedback distortion from capturing the speaker output along with user input
Solution Approach 1:
The system extracts and separates the user's voice signal from the mixed audio signal that contains both the speaker device output and user input. By using voice activity detection and signal processing techniques, the system isolates the user's utterance from the feedback path, removing the harmful speaker output component while preserving the user's speech for accurate recognition
Solution Approach 2:
The system introduces an intermediary processing stage between the speaker device and the speech recognition engine. This intermediary component analyzes the audio signal, identifies user speech portions, and selectively processes them for recognition while filtering out the speaker device output, thereby mediating between the conflicting requirements of maintaining audio output and achieving recognition accuracy
2Measurement precision
If the audio output level of the speaker device is reduced to suppress background audio, then the speech recognition accuracy improves, but the audio output quality deteriorates
Solution Approach 1:
The system dynamically adjusts the speaker device audio output level based on real-time detection of user speech activity. When user speech is detected, the system temporarily reduces the speaker output level to minimize feedback distortion; when no user speech is present, the system restores normal audio output levels. This dynamic adjustment allows the system to optimize both audio output quality and speech recognition accuracy at different moments in time
3Speed
If the system captures audio continuously to recognize user input, then the response time to user queries improves, but the distortion from speaker feedback increases
Solution Approach 1:
The system implements a feedback mechanism where the captured audio signal is analyzed to detect user speech activity, and this detection information is used to control the speaker device output level. The system continuously monitors the audio input, identifies when user speech is present, and automatically adjusts the speaker output accordingly, creating a closed-loop control system that responds to user input while minimizing feedback distortion
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.