Distributed Speech Recognition via Loose Coupling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic speech recognition systems face challenges in providing a balance between recognition quality, speed, and cost, especially with the increasing variety of computing devices that require improved techniques for distributed speech recognition capabilities.
Innovation Solution
The system separates the audio capture component from the result processing component, allowing them to reside on different physical and logical devices while maintaining a dynamic, loose coupling through context sharing, enabling efficient communication and processing of speech recognition results across distributed components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If audio capture and result processing are performed on the same device, then system simplicity is maintained, but speech recognition quality and processing efficiency deteriorate due to suboptimal audio signal processing
Solution Approach 1:
The system divides the speech recognition functionality into separate components: audio capture component on one device and result processing component on another device. This segmentation allows each component to be optimized for its specific function, improving speech recognition quality while managing system complexity through modular architecture
Solution Approach 2:
The patent introduces a context sharing component as an intermediary that facilitates communication between the audio capture component and result processing component. This mediator enables loose coupling while maintaining coordinated operation, resolving the contradiction between separation benefits and system complexity
2Measurement precision
If speech recognition is processed locally on the device, then response speed is fast, but recognition quality deteriorates due to limited processing capabilities
Solution Approach 1:
By separating audio capture from result processing and distributing them across different devices, the system enables specialized processing. The result processing component can use more powerful computational resources for high-quality recognition while the audio capture component responds immediately to user speech
Solution Approach 2:
The system performs preliminary audio capture and transmission before full processing occurs. The audio capture component prepares and sends audio data to the result processing component, allowing processing to begin immediately upon receipt without waiting for complete audio input
3Adaptability or versatility
If a unified speech recognition system is used across various devices, then system simplicity is maintained, but adaptability to different devices and contexts deteriorates
Solution Approach 1:
The patent creates a universal speech recognition architecture where the audio capture component and result processing component can operate independently on different device types. The context sharing mechanism enables this same architecture to adapt to various devices (smartphones, tablets, computers, wearables) without requiring device-specific implementations
Solution Approach 2:
The system employs dynamic context sharing that adapts to different devices and usage scenarios. The context information exchanged between components can be customized based on device capabilities, user preferences, and environmental factors, enabling versatile adaptation across diverse platforms
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
An automatic speech recognition system includes an audio capture component, a speech recognition processing component, and a result processing component which are distributed among two or more logical devices and/or two or more physical devices. In particular, the audio capture component may be located on a different logical device and/or physical device from the result processing component. For example, the audio capture component may be on a computer connected to a microphone into which a user speaks, while the result processing component may be on a terminal server which receives speech recognition results from a speech recognition processing server.