In-Cloud Wake-Up System Offloads Voice Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wake-up solutions for smart devices require high power consumption and resource occupation, leading to potential quality degradation due to the need for memory compression.
Innovation Solution
An in-cloud wake-up method and system that receives voice input, performs decoding using an acoustic model, language model, and pronunciation dictionary in the cloud, identifies wake-up characters, and generates instructions to reduce resource consumption and power loss while maintaining wake-up quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If local wake-up model is used in terminal device, then wake-up function can be implemented, but power consumption and resource occupation increase
Solution Approach 1:
The patent extracts the wake-up model from the terminal device and places it in the cloud server. The terminal device only needs to upload voice data and receive instructions, while the computationally intensive wake-up model runs remotely on the cloud server, thereby reducing local power consumption and resource occupation while maintaining wake-up quality.
Solution Approach 2:
The patent introduces the cloud server as an intermediary between the terminal device and the wake-up model. The cloud server acts as a mediator that receives voice data from the terminal, processes it through the wake-up model, and returns wake-up instructions, thereby offloading computational burden from the terminal device.
2Quantity of substance
If wake-up model is compressed to fit terminal memory, then memory space constraint is satisfied, but wake-up effect deteriorates and wake-up mistakes increase
Solution Approach 1:
The patent extracts the wake-up model from the terminal device memory constraints and places it in the cloud server with充足的存储空间. This allows the wake-up model to be stored in full capacity without compression, maintaining high recognition accuracy while the terminal device only needs to store minimal local data.
3Productivity
If wake-up model is stored locally in terminal device, then wake-up processing can be done locally, but resource occupation increases
Solution Approach 1:
The patent extracts the resource-intensive wake-up model from the terminal device and relocates it to the cloud server. The terminal device performs only lightweight operations (voice upload and instruction reception), while the cloud server handles the computationally intensive wake-up model processing, thereby reducing local resource occupation.
Data Source
AI summary
An in-cloud wake-up method and system, a terminal and a non-volatile computer-readable storage medium are provided. The in-cloud wake-up method includes: receiving wake-up voice input by a user, and transmitting the wake-up voice to cloud; performing an decoding operation on the wake-up voice in cloud to generate a wake-up text; identifying wake-up characters in the wake-up text; and providing a wake-up instruction according to an identification result. With the in-cloud wake-up method, voice can be identified in cloud, and a smart device can be waken up in cloud according to the identified voice, so that resource consumption and power loss of the terminal device are reduced in the case of ensuring the quality of wake-up.