AI Speech Wake-Up Network Using Garbage Word List
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech wake-up technologies face challenges in accuracy and false alarm rates, particularly in dynamic environments, and require manual intervention and significant memory resources, limiting their application in portable devices.
Innovation Solution
The method constructs a network for identifying wake-up words using a preset garbage word list and approximate pronunciation information, allowing for dynamic network creation for customized wake-up words, reducing false alarms, and optimizing for low power consumption and memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional language models are used for speech wake-up recognition, then recognition coverage is improved, but memory resource consumption increases and false alarm rate increases
Solution Approach 1:
The patent extracts and removes the language model component from the speech recognition system, replacing it with a garbage word list. This extraction eliminates the high memory consumption associated with language models while maintaining wake-up recognition functionality through a simplified approach that focuses only on distinguishing wake-up words from garbage sounds.
Solution Approach 2:
The patent replaces the expensive, memory-intensive language model with a cheap, lightweight garbage word list that can be easily stored and processed. This substitution uses minimal memory resources while effectively identifying wake-up words, making the system suitable for portable devices with limited resources.
2Ease of operation
If traditional speech wake-up technology is used, then basic wake-up functionality is achieved, but false alarm rate increases in dynamic environments
Solution Approach 1:
The patent performs preliminary classification by categorizing input sounds as either garbage words or potential wake-up words before conducting detailed recognition. This preliminary action filters out background noise and irrelevant sounds, significantly reducing false alarms in dynamic environments while maintaining ease of wake-up operation.
Solution Approach 2:
The patent segments the speech recognition process into distinct stages: garbage word detection, wake-up word identification, and command processing. This segmentation allows the system to handle different types of inputs appropriately, reducing false alarms by preventing garbage sounds from triggering wake-up operations.
3Measurement precision
If manual intervention is required for wake-up operations, then control precision is improved, but operation complexity increases and automation decreases
Solution Approach 1:
The patent implements a self-service wake-up system where the device automatically detects and processes wake-up words without requiring manual intervention. The system autonomously distinguishes between garbage sounds and valid wake-up commands, providing both high automation and precise control through the garbage word list filtering mechanism.
Solution Approach 2:
The patent incorporates feedback mechanisms where the system continuously monitors input sounds, compares them against the garbage word list, and adjusts its recognition behavior based on the results. This feedback loop enables automatic wake-up operations with high precision by learning from and adapting to different acoustic environments.
Data Source
AI summary
A method and a device for waking up via a speech based on artificial intelligence are provided in the present disclosure. The method includes: acquiring pronunciation information of a customized wake-up word; acquiring approximate pronunciation information of the pronunciation information; and constructing a network for identifying wake-up words according to a preset garbage word list, the pronunciation information and the approximate pronunciation information, identifying an input speech according to the network to acquire an identified result, and determining whether to perform a wake-up operation according to the identified result. With embodiments of the present disclosure, different networks for identifying the wake-up words may be constructed dynamically for different customized wake-up words, thus effectively improving an accuracy of waking up, reducing a false alarm rate, improving an efficiency of waking up, occupying less memory, and having low power consumption.