Audio Data Processing Decoupling Keyword Detection and Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio processing technologies often integrate speech detection and recognition into a single application, which can lead to inefficiencies when multiple applications require separate speech detection and recognition capabilities.
Innovation Solution
A method where a first speech application performs keyword detection on audio data to obtain a key frame, determines a second speech application based on the key frame, and sends a frame identifier to the second application for speech recognition, allowing for decoupling of keyword detection and speech recognition into separate applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If speech detection and recognition are integrated into the same application, then the system structure is simple, but the configurability and flexibility are poor when multiple applications require separate speech processing capabilities
Solution Approach 1:
The patent segments speech processing into two independent applications: a first speech application for keyword detection and a second speech application for speech recognition. This segmentation allows each application to be independently configured and activated based on specific needs, thereby improving adaptability and versatility without significantly complicating the overall system structure.
2Adaptability or versatility
If speech detection and recognition are performed by separate applications, then the configurability improves, but the device complexity increases
Solution Approach 1:
The patent implements a universal audio data sharing mechanism where the first speech application can provide audio data to multiple second speech applications through a standardized interface. This multi-functionality allows the system to maintain high configurability while reducing complexity by reusing the same audio data processing pipeline across different applications.
3Reliability
If multiple applications perform full speech recognition independently, then each application has complete functionality, but the computing power requirements increase significantly
Solution Approach 1:
The patent extracts the audio data processing and keyword detection functions into a separate first speech application, leaving the second speech application to focus only on speech recognition. This extraction allows the system to maintain complete functionality while reducing computing power requirements by avoiding redundant audio processing in multiple applications.
4Device complexity
If keyword detection is performed by the speech recognition application, then the system is unified, but the speech recognition efficiency decreases when multiple applications are involved
Solution Approach 1:
The patent segments the speech processing pipeline so that keyword detection is performed by the first speech application and speech recognition is performed by the second speech application. This segmentation enables parallel processing and specialized optimization of each function, thereby improving speech recognition efficiency when multiple applications are involved while maintaining system unity through standardized interfaces.
Data Source
AI summary
A method for processing audio data includes: obtaining audio data, and obtaining a key frame of the audio data by performing keyword detection on the audio data; determining a second speech application based on the key frame, and sending a frame identifier of the key frame to the second speech application; and extracting first audio data from the audio data according to the key frame, and sending the first audio data to the second speech application for speech recognition.