Edge Speech Recognition Using Distributed NLU Model Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition technologies face limitations in real-time processing due to reliance on cloud-based systems, which can be hindered by unstable communication connections and poor performance during communication failures, and struggle to efficiently recognize frequently used voice commands.
Innovation Solution
Implementing a distributed network system that utilizes an edge computing device to perform speech recognition by receiving and processing voice data using a personalized natural language understanding (NLU) model generated in the cloud, allowing for real-time processing and reducing reliance on cloud connectivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speech recognition is performed in a cloud environment, then speech recognition accuracy can be maintained, but real-time processing capability deteriorates due to communication transmission requirements
Solution Approach 1:
The system divides speech recognition processing into two segments: cloud-based NLU model generation for accuracy, and edge-based ASR inference for real-time speed. The NLU model is trained in the cloud using comprehensive data, then deployed to edge devices for local real-time speech processing, separating the training and inference phases to optimize both accuracy and speed.
Solution Approach 2:
The edge computing device serves as an intermediary between the cloud server and the speech recognition system. It receives the NLU model from the cloud, performs local ASR processing, and only transmits results or requests when necessary, reducing communication overhead while maintaining cloud-based model accuracy.
2Adaptability or versatility
If speech recognition relies on cloud connection, then comprehensive processing capability is achieved, but system reliability deteriorates due to communication instability
Solution Approach 1:
The system performs preliminary action by pre-training the NLU model in the cloud and caching it locally on the edge device before actual speech recognition is needed. This allows the edge device to operate independently during communication failures, ensuring reliability while maintaining comprehensive processing capability through the pre-loaded model.
Solution Approach 2:
The system implements local quality by enabling the edge device to perform autonomous speech recognition processing locally using the cached NLU model. This local processing capability ensures continuous operation during cloud connectivity issues, while the cloud connection provides enhanced capabilities when available, creating a reliable hybrid system.
3Extent of automation
If all speech recognition processing is performed in cloud, then centralized control is maintained, but network overhead increases and real-time performance deteriorates
Solution Approach 1:
The system extracts the inference function from the cloud environment and places it on the edge device. The cloud server retains only the model training and distribution functions, while the edge device performs local speech recognition inference, eliminating the need for continuous cloud communication during speech processing and significantly reducing network overhead.
Data Source
AI summary
Disclosed herein is a speech recognition method in a distributed network environment. A method of performing a speech recognition operation in an edge computing device includes receiving a natural language understanding (NLU) model from the cloud server, storing the received NLU model, receiving voice data spoken by a user from the client device, performing a natural language processing operation on the received voice data using the NLU model, performing speech recognition according to the natural language processing operation, and transmitting a result of the speech recognition to the client device.At least one of the edge computing device, a voice recognition device, and a server may be associated with an artificial intelligence module, a drone (an unmanned aerial vehicle (UAV)), a robot, an augmented reality (AR) device, a virtual reality (VR) device, a device related to a 5G service, and the like.