Hotphrase Detection Through Sequential Hotwords for Low-Power Flexibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional low-powered hotphrase detection models lack flexibility, requiring precise commands and failing to accommodate variations in user phrasing, limiting the range of acceptable commands in speech-enabled environments.
Innovation Solution
Implement a two-stage hotphrase detection system where a first stage detects multiple hotwords in an utterance and aggregates them to identify a full hotphrase, followed by a second-stage validation using automated speech recognition to confirm the presence of the hotphrase, allowing for flexible and natural user interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If conventional low-powered hotphrase detection models are used, then power consumption is reduced, but flexibility and acceptance of varied user phrasing deteriorate
Solution Approach 1:
The system segments hotphrase detection into two distinct stages: a first stage using low-powered detection models to identify individual hotwords, and a second stage combining multiple hotword detections to recognize complete hotphrases. This segmentation allows the system to use computationally efficient models while achieving flexible hotphrase recognition through the combination of multiple detections.
Solution Approach 2:
The system performs preliminary detection of individual hotwords using low-powered models before committing to full hotphrase recognition. By detecting hotwords in advance and aggregating them according to sequential rules, the system prepares potential hotphrase candidates without requiring expensive processing for every audio input, thus maintaining flexibility while conserving power.
2Measurement precision
If precise command phrasing is required, then detection accuracy is improved, but ease of operation and user interaction flexibility deteriorate
Solution Approach 1:
The system dynamically adjusts its recognition criteria based on the aggregation of multiple hotword detections. Rather than requiring exact phrase matching, the system allows flexible combinations of detected hotwords to form valid hotphrases, adapting to various user phrasing styles while maintaining detection accuracy through sequential detection rules and time-window constraints.
Solution Approach 2:
The system changes the detection parameter from requiring exact phrase matching to accepting sequential combinations of individual hotwords. By modifying the recognition criterion from fixed phrase matching to flexible hotword aggregation, the system improves ease of operation while maintaining detection precision through the structured combination rules.
3Measurement precision
If full automated speech recognition is performed continuously, then speech recognition accuracy is improved, but power consumption and processing cost increase
Solution Approach 1:
The system performs partial speech recognition by using low-powered hotword detection models for initial filtering, applying full automated speech recognition only when multiple hotwords are detected in sequence. This partial application of expensive processing only when necessary maintains speech recognition accuracy for actual hotphrases while dramatically reducing overall power consumption compared to continuous full ASR.
Solution Approach 2:
The system performs preliminary filtering using low-powered hotword detection before engaging full automated speech recognition. By detecting and aggregating hotwords in advance, the system identifies candidate hotphrases that warrant full ASR processing, thereby reducing the frequency and power consumption of expensive speech recognition operations while maintaining accuracy for valid commands.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method (300) includes receiving audio data (202) corresponding to an utterance (110) captured by a user device (104). The utterance includes a command for a digital assistant (109) to perform an operation. The method also includes determining, using a hotphrase detector (200) configured to detect each trigger word in a set of trigger words associated with a hotphrase (210), whether any of the trigger words in the set of trigger words are detected in the audio data during a corresponding fixed-duration time window (220). The method also includes identifying, in the audio corresponding to the utterance, the hotphrase when each other trigger word in the set of trigger words was also detected in the audio data. The method also includes triggering an automated speech recognizer (116) to perform speech recognition on the audio data when the hotphrase is identified in the audio data corresponding to the utterance.