Wake Word Verification for Multi-Assistant Voice Request Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users interacting with multiple voice assistants face challenges such as resource consumption, incorrect activation, and privacy concerns due to limited device resources and increased likelihood of mistakes, especially when managing multiple voice assistants on a single device.
Innovation Solution
A system comprising a voice assistant manager that detects a wake word in a user utterance, identifies the appropriate voice assistant, and verifies the wake word multiple times through the voice assistant and a cloud service to ensure accurate processing and resource management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple voice assistants are activated on a single device, then the system provides more functionality and versatility, but resource consumption increases and the likelihood of false activations increases
Solution Approach 1:
The patent segments the wake word detection process across multiple independent components: the voice assistant manager performs initial detection, each voice assistant performs verification detection, and the cloud service performs final verification. This segmentation allows the system to distribute resource consumption across multiple specialized modules rather than having all voice assistants continuously active, thereby maintaining versatility while managing resource usage efficiently.
Solution Approach 2:
The voice assistant manager performs preliminary wake word detection before activating any voice assistant. This preliminary action filters out false positives early in the process, preventing unnecessary activation of voice assistants and their associated resource consumption. The system only proceeds to full voice assistant activation when the manager confidently detects a wake word, thus conserving resources while maintaining functionality.
2Adaptability or versatility
If multiple voice assistants are activated on a single device, then the system provides more functionality, but the likelihood of incorrect activation increases
Solution Approach 1:
The patent implements a feedback mechanism where the detected wake word is verified by multiple independent components (voice assistant manager, voice assistant, and cloud service). Each component provides feedback on whether the wake word was correctly detected, and the system only activates the voice assistant when all feedback confirms successful detection. This multi-layer feedback system significantly reduces false activations while maintaining the ability to support multiple voice assistants.
Solution Approach 2:
The system prepares multiple verification layers in advance to cushion against potential detection errors. Before any voice assistant activation occurs, the wake word detection is cushioned by redundant verification at the voice assistant level and cloud service level. This beforehand cushioning ensures that even if one detection component makes an error, the other components will catch and correct it, thereby maintaining high activation accuracy.
3Reliability
If wake word detection is performed multiple times, then false positives are reduced and activation accuracy is improved, but processing time increases
Solution Approach 1:
The patent implements periodic action by structuring the wake word detection as a sequence of discrete verification stages rather than continuous processing. The voice assistant manager performs detection periodically, then the system moves to the next verification stage only when needed. This periodic structure allows the system to achieve high reliability through multiple verifications while minimizing processing time by only executing subsequent verification stages when the previous stage indicates a potential wake word detection.
4Use of energy by moving object
If the voice assistant manager detects and routes to a specific voice assistant, then resource usage is optimized, but the risk of sending information to the wrong assistant increases
Solution Approach 1:
The patent introduces the voice assistant manager as an intermediary between wake word detection and voice assistant activation. The manager performs initial detection and routing decisions based on the detected wake word, thereby optimizing resource usage by activating only the relevant voice assistant. The intermediary role includes a verification step where the manager confirms the detected wake word matches the intended voice assistant before routing, thus maintaining high routing accuracy while achieving resource optimization.
Data Source
AI summary
A system for processing voice requests includes a voice assistant manager and a plurality of voice assistants. The voice assistant manager detects a wake word in an utterance and communicates the utterance to a voice assistant of the plurality of voice assistants. In some embodiments, the voice assistant may verify the detected wake word and communicate with a cloud service, which may also verify the detected wake word and generate a response to the utterance. In some embodiments, the voice assistant manager may activate or deactivate one or more of the voice assistants.