Wake-Up Word Recognition Using Speaker Identity in Multi-Speaker Audio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech-enabled systems struggle with robust wake-up word detection in environments with multiple speakers, leading to unsatisfactory responsiveness and user frustration, particularly when hands-free interaction is desired.
Innovation Solution
A speech-enabled system that utilizes speaker identity information to personalize wake-up words and improve detection accuracy by using personalized lists of wake-up words associated with each speaker, and employs acoustic analysis to differentiate between speakers and adapt recognition components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a speech-enabled system uses a single wake-up word for all users, then the system is simple to operate, but the detection accuracy decreases in environments with multiple speakers
Solution Approach 1:
The patent segments the wake-up word detection system by creating speaker-specific wake-up words and personalized lists for each user. The system divides the acoustic environment analysis into speaker-specific segments, allowing each user to have customized wake-up words that are recognized only when spoken by them. This segmentation resolves the contradiction by improving detection accuracy through personalization while maintaining operational simplicity for each individual user.
Solution Approach 2:
The system dynamically adapts wake-up word recognition based on identified speaker characteristics. It dynamically creates and updates personalized wake-up word lists for each speaker, adjusting the recognition parameters according to the specific acoustic patterns and characteristics of each user. This dynamic adaptation improves detection accuracy without requiring complex manual configuration from users.
2Productivity
If the system processes all acoustic input regardless of speaker identity, then no speaker information is lost, but the system responsiveness and accuracy decrease
Solution Approach 1:
The system performs preliminary speaker identification and characterization before wake-up word recognition. It pre-processes acoustic input to identify speaker characteristics and creates speaker-specific wake-up word lists in advance. This preliminary action allows the system to quickly respond to wake-up words from authorized speakers while filtering out irrelevant acoustic input, thereby improving responsiveness without losing critical speaker identity information.
Solution Approach 2:
The patent extracts and isolates speaker identity information from the acoustic environment using speaker identification techniques. By separating speaker characteristics from general acoustic input, the system can focus processing resources on recognizing wake-up words from identified speakers. This extraction improves system responsiveness by eliminating unnecessary processing of non-speaker-specific acoustic data while preserving essential speaker identity information.
3Ease of operation
If manual triggers are used to activate speech recognition, then the system can control when processing occurs, but user convenience decreases when hands are occupied
Solution Approach 1:
The system uses speaker identification feedback to confirm that a wake-up word is genuinely intended for the system. By continuously monitoring speaker characteristics and comparing them against stored speaker profiles, the system provides feedback to verify authentic wake-up word usage. This feedback mechanism reduces false positives while maintaining hands-free operation, as the system can reliably distinguish between intentional wake-up words and accidental similar-sounding inputs.
Solution Approach 2:
The patent replaces manual mechanical triggers with acoustic-based speaker identification and recognition. Instead of requiring physical button presses or manual activation, the system uses voice-based speaker verification to automatically determine when speech processing should occur. This substitution achieves hands-free operation while maintaining reliability through accurate speaker identification and contextual analysis of the acoustic environment.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system for detection of at least one designated wake-up word for at least one speech-enabled application. The system comprises at least one microphone; and at least one computer hardware processor configured to perform: receiving an acoustic signal generated by the at least one microphone at least in part as a result of receiving an utterance spoken by a speaker; obtaining information indicative of the speaker's identity; interpreting the acoustic signal at least in part by determining, using the information indicative of the speaker's identity and automated speech recognition, whether the utterance spoken by the speaker includes the at least one designated wake-up word; and interacting with the speaker based, at least in part, on results of the interpreting.