Artificial intelligence device for providing speech recognition function and method of operating artificial intelligence device
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI devices experience delays in responding to user commands due to the need to transmit all commands to a natural language processing server, especially when multiple wake-up words are used, leading to memory inefficiency and slowed speech recognition performance.
Innovation Solution
An AI device employing a basic wake-up word for activating speech recognition and an additional wake-up word for controlling operations, allowing the device to perform actions without server intervention in specific situations, thereby reducing the need for memory storage of all operation commands and enhancing speech recognition speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all additional commands are transmitted to the NLP server for speech control, then comprehensive speech recognition is achieved, but delay occurs in data exchange process causing user inconvenience
Solution Approach 1:
The patent segments wake-up words into two categories: basic wake-up words stored in cloud servers for comprehensive recognition, and additional wake-up words stored locally in the artificial intelligence device for immediate response. This segmentation allows the system to handle different types of commands through different paths, resolving the contradiction between comprehensive recognition and fast response.
2Adaptability or versatility
If multiple wake-up words are stored in the artificial intelligence device, then speech control versatility is improved, but memory capacity is limited or computation for recognizing wake-up word increases
Solution Approach 1:
The patent divides wake-up word storage between cloud servers (basic wake-up words) and local device memory (additional wake-up words). This segmentation enables the device to support multiple wake-up words without overwhelming local memory capacity, while still providing comprehensive speech control capabilities.
Solution Approach 2:
The patent introduces cloud servers as an intermediary for storing and processing basic wake-up words, freeing up local device memory. The cloud server acts as a mediator that handles the storage burden, allowing the device to maintain versatility with limited local resources.
3Ease of operation
If basic wake-up word recognition activates speech recognition function, then speech control is enabled, but all commands require server transmission causing operational delays
Solution Approach 1:
The patent segments command processing into two paths: commands containing additional wake-up words are processed locally for immediate execution, while other commands are transmitted to the server. This segmentation enables fast local response for specific commands while maintaining comprehensive server processing for others.
Solution Approach 2:
The patent performs preliminary action by storing additional wake-up words locally in the device before they are needed. This pre-storation enables the device to immediately recognize and execute these wake-up words without waiting for server communication, reducing command execution time.
Data Source
AI summary
An artificial intelligence device for providing a speech recognition function includes a memory configured to store a basic wake-up word used to activate the speech recognition function of the artificial intelligence device and an additional wake-up word used to control operation of the artificial intelligence device, a microphone configured to receive a speech command, and a processor configured to determine whether a current situation is an additional wake-up word recognition situation when the basic wake-up word is recognized from the speech command and perform operation corresponding to the remaining command excluding the basic wake-up word from the speech command upon determining that the current situation is the additional wake-up word recognition situation.