Multi-Assistant Trigger Word Detection Using Confidence Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio responsive devices struggle with selecting the correct digital assistant from multiple digital assistants due to unreliable voice input recognition, echo interference, and poor audio output quality, especially with low-quality microphones and speakers.
Innovation Solution
Implementing a system that detects trigger words, calculates echo cancellation delay, and dynamically adjusts equalization based on volume level to enhance voice input processing and output quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple digital assistants are supported with trigger word detection, then the versatility of the device is improved, but the reliability of selecting the correct digital assistant deteriorates due to false detections from accents and low-quality microphones
Solution Approach 1:
The patent divides the trigger word detection task into separate modules, with each digital assistant having its own dedicated trigger word detection module. This segmentation allows each module to be optimized for its specific assistant's trigger words, reducing cross-assistant interference and improving overall detection accuracy while maintaining support for multiple assistants.
Solution Approach 2:
The patent applies local quality by customizing the trigger word detection parameters and thresholds for each digital assistant based on their specific trigger words and acoustic characteristics. This allows each assistant to have optimized detection settings that account for variations in accents, microphone quality, and environmental factors, thereby improving reliability while maintaining versatility.
2Reliability
If echo cancellation is implemented in non-real-time operating systems, then audio quality is improved, but the system complexity increases due to inability to predict echo delay
Solution Approach 1:
The patent performs preliminary action by measuring and storing echo delay characteristics during system initialization or in real-time when conditions are favorable. These pre-measured delay values are then used for echo cancellation processing, eliminating the need for complex real-time prediction algorithms and reducing system complexity while maintaining audio quality.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously monitors audio output and adjusts echo cancellation parameters based on detected echo patterns. This feedback approach allows the system to adapt to changing acoustic environments without requiring complex predictive models, thereby improving audio quality while managing system complexity through iterative adjustment rather than sophisticated prediction.
3Reliability
If audio equalization is applied to compensate for speaker non-linearities, then audio output quality is improved, but the adaptability to different volume levels deteriorates as fixed equalization cannot dynamically adjust
Solution Approach 1:
The patent applies dynamics by making the audio equalization parameters variable rather than fixed. The system dynamically adjusts equalization settings based on the current volume level being used for playback. This allows the equalization to compensate for speaker non-linearities at different operating points, maintaining audio quality across a wide range of volume levels while adapting to user preferences and environmental conditions.
Data Source
AI summary
Disclosed herein are system, apparatus, article of manufacture, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for selecting a digital assistant from among multiple digital assistants. An embodiment operates by receiving a voice input containing a trigger word at a first voice adapter associated with a digital assistant that generates a first confidence score for the trigger word. The embodiment further receives the voice input at a second voice adapter that generates a second confidence score for the trigger word. The embodiment determines the first confidence score is higher than the second confidence score. The embodiment selects the digital assistant based on the determining.