Personalized Wake-Up Keyword Detection for Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems face challenges in accurately activating the speech recognition function, especially when multiple devices with the same speech recognition feature are in proximity, leading to unintended activation and errors in recognizing voice commands.
Innovation Solution
A system comprising a device and a speech recognition server that utilize a personalized wake-up keyword model to detect and differentiate user voice commands, with the device detecting the wake-up keyword and transmitting signals to the server for accurate recognition and processing, ensuring reliable activation and error reduction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a fixed wake-up keyword is used to activate speech recognition, then the activation process is simple, but unintended devices may be activated and recognition errors occur
Solution Approach 1:
The patent changes the parameters of the wake-up keyword from a fixed universal keyword to a personalized keyword that is unique to each user-device pair. This is achieved by generating personalized wake-up keywords through voice printing technology that captures individual voice characteristics, thereby maintaining activation simplicity while significantly improving activation accuracy and preventing unintended device activation.
2Device complexity
If wake-up keyword and voice command are processed separately, then the processing flow is clear, but activation errors occur when keywords and commands are input simultaneously
Solution Approach 1:
The patent merges the wake-up keyword detection and voice command recognition into a unified continuous processing framework. Instead of treating them as separate sequential steps, the system continuously performs voice printing and wake-up keyword detection, allowing simultaneous input of keywords and commands to be processed together, thereby eliminating activation errors while maintaining clear processing logic.
3Adaptability or versatility
If multiple devices with speech recognition are in proximity, then device availability increases, but unintended device activation occurs
Solution Approach 1:
The patent applies local quality by making each device's wake-up keyword recognition specific to its registered user's voice characteristics. Through voice printing technology, each device learns and stores the unique voice features of its authorized user, creating a localized recognition zone. This ensures that when multiple devices are in proximity, only the intended device responds to its owner's voice, maintaining high device availability while ensuring precise device specificity.
Data Source
AI summary
A device detects a wake-up keyword from a received speech signal of a user by using a wake-up keyword model, and transmits a wake-up keyword detection/non-detection signal and the received speech signal of the user to a speech recognition server. The speech recognition server performs a recognition process on the speech signal of the user by setting a speech recognition model according to the detection or non-detection of the wake-up keyword.