Multi-Device Wakeword Detection Using Confidence-Based Audio Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for locating audio devices in a home environment often require a single device to be chosen for audio processing, leading to suboptimal performance due to potential misidentification of the source of a wakeword, especially when users are near boundaries between zones.
Innovation Solution
A method for selecting a device for audio processing based on comparing wakeword confidence metrics from multiple devices, determining local maxima in wakeword confidence values, and choosing the most suitable device for subsequent audio processing, such as speech recognition or command recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single device is chosen for audio processing in a multi-device environment, then device complexity is reduced, but measurement precision of wakeword source location deteriorates
Solution Approach 1:
The patent segments the wakeword detection task across multiple devices, with each device independently detecting wakewords in its local zone and reporting confidence metrics. This segmentation allows the system to maintain low individual device complexity while achieving high overall measurement precision through collective detection and comparison of wakeword confidence metrics from multiple spatially distributed devices.
2Measurement precision
If multiple devices perform wakeword detection, then measurement precision of wakeword source location improves, but device complexity increases
Solution Approach 1:
The patent merges the detection capabilities of multiple devices while maintaining independent operation. Each device performs local wakeword detection and confidence metric generation independently, then these results are combined by selecting the device with the highest confidence metric. This merging approach achieves high measurement precision through multiple detectors while controlling system complexity by avoiding centralized processing of raw audio data.
3Device complexity
If a device far from the user triggers on a wakeword, then device complexity remains low with simple threshold detection, but reliability of audio processing deteriorates
Solution Approach 1:
The patent replaces simple mechanical threshold-based detection with a confidence metric comparison mechanism. Instead of relying solely on fixed thresholds that may trigger falsely on distant devices, the system uses dynamic confidence metrics that reflect the quality and strength of wakeword detection. This substitution maintains relatively simple device-level operation while significantly improving reliability through intelligent selection based on comparative confidence assessment.
Data Source
AI summary
A method for selecting a device for audio processing may involve receiving a first wakeword confidence metric from a first device that includes at least a first microphone and receiving a second wakeword confidence metric from a second device that includes at least a second microphone. The first and second wakeword confidence metrics may correspond to a first local maximum of a first plurality of wakeword confidence values determined by the first device and a second local maximum of a second plurality of wakeword confidence values determined by the second device. The method may involve comparing the first wakeword confidence metric and the second wakeword confidence metric and selecting a device for subsequent audio processing based, at least in part, on a comparison of the first wakeword confidence metric and the second wakeword confidence metric.