Hotword Detection Soft Accept Logic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current hotword detection systems operate with a binary approach, leading to false accepts and false rejects due to the inability to account for borderline confidence cases, resulting in incorrect interpretations and user frustration.
Innovation Solution
A non-binary approach is introduced, incorporating a 'soft accept' outcome category, where the system detects a hotword with lower confidence and prompts the user for affirmative acknowledgement before performing the action, allowing for a limited response that reflects the confidence level and reduces intrusive errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a binary hotword detection approach is used, then the system can operate with simple decision logic, but it produces false accepts and false rejects due to inability to handle borderline confidence cases
Solution Approach 1:
The hotword detection outcome is segmented into three distinct categories: hard accept, soft accept, and reject. This segmentation allows the system to differentiate between high-confidence detections (hard accept), borderline detections (soft accept requiring user confirmation), and low-confidence detections (reject), thereby improving detection accuracy without significantly increasing system complexity.
Solution Approach 2:
The system changes the parameter of detection confidence by introducing a threshold-based classification system. Instead of a single binary threshold, two thresholds are used to create three outcome categories, allowing the system to adapt its response based on the confidence level of the detection.
2Speed
If the system triggers action immediately upon hotword detection, then the system responds quickly to user commands, but it causes user frustration due to incorrect interpretations from false accepts
Solution Approach 1:
For soft accept cases, the system performs a preliminary action by presenting the detected hotword and requested action to the user for confirmation before executing the action. This preliminary confirmation step prevents incorrect interpretations from causing user frustration, while still maintaining quick response times for hard accept cases.
Solution Approach 2:
The user is introduced as an intermediary between hotword detection and action execution for soft accept cases. The user's confirmation serves as a mediator that validates borderline detections before triggering actions, reducing false accepts while maintaining system responsiveness.
3Reliability
If the system requests user confirmation for all hotword detections, then false accepts are minimized, but the system becomes intrusive and reduces productivity
Solution Approach 1:
The system applies different quality levels of confirmation based on the detection confidence. Hard accept cases (high confidence) proceed directly to action execution without user confirmation, maintaining productivity. Soft accept cases (borderline confidence) request user confirmation, ensuring reliability. This local differentiation of confirmation requirements optimizes both productivity and reliability.
Data Source
AI summary
A method for a soft acceptance of a hotword receives audio data characterizing a soft hotword event detected by a hotword detector in streaming audio captured by a user device. The method also processes the audio data to determine that the audio data corresponds to a query specifying an action to perform on the user device. Without triggering performance of the action on the user device or the other device, the method provides a notification for output from the user device where the notification prompts a user associated with the user device to provide an affirmative input indication in order to trigger performance of the action on the user device or the other device and, when the user fails to provide the affirmative input indication, instructs the user device or the other device to not perform the action specified by the query.