Vehicular Voice Recognition System Intent Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice recognition systems for vehicles struggle to accurately infer user intentions, especially when external noise is present or when spoken instructions are incomplete, leading to unsatisfactory user experiences due to incorrect content provision or service interruptions.
Innovation Solution
A vehicular voice recognition system that performs multiple inference operations based on user instruction history, service use patterns, and preferred items to determine the appropriate service domain and provide relevant services, even when instructions are not perfectly defined, using a controller that converts input instructions to text and utilizes databases to suggest services through audio and image outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If a web-based search is performed when the user's intent is not grasped, then the system may provide contents to decipher the user's intent, but the system may provide contents that do not meet the user's intent or may interrupt a voice service due to error
Solution Approach 1:
The system performs preliminary analysis of the spoken sentence structure and keywords before initiating web-based search. It checks whether the sentence meets predefined completion criteria and contains sufficient semantic information. Only when these preliminary conditions are satisfied does the system proceed to web search, preventing premature or erroneous searches that would interrupt voice services.
Solution Approach 2:
The system implements a feedback mechanism where the results of web-based search are evaluated against the original user input and context. If the search results do not align with the detected user intent or fail to provide meaningful information, the system terminates the search and maintains the original voice service flow, thereby preventing erroneous content delivery.
2Measurement precision
If the system only provides voice service when the defined form and structure of the spoken sentence is completed, then the system can ensure accurate recognition, but if the user does not know how to use or fails to speak the finished sentence, the voice service fails
Solution Approach 1:
The system applies partial matching rules that allow voice service activation even when the spoken sentence does not fully match the predefined complete structure. It identifies key keywords and semantic elements within incomplete sentences and attempts to provide appropriate services based on partial information, thereby reducing the burden on users to speak perfectly formed sentences.
Solution Approach 2:
The system dynamically adjusts the required sentence completion threshold based on context, user history, and the detected confidence level of keyword matching. When the user demonstrates difficulty with complete sentence formation, the system adapts by accepting shorter or less structured inputs, making the voice service more accessible while maintaining reasonable accuracy.
3Loss of information
If the system performs multiple inference operations based on stored data, then the system can infer user intentions more accurately, but the system complexity increases with multiple inference operations
Solution Approach 1:
The inference system is segmented into multiple independent modules, each responsible for a specific type of inference operation (e.g., keyword-based inference, history-based inference, pattern-based inference). Each module processes specific aspects of user input independently and outputs results that are integrated by a coordinator. This modular structure manages complexity by dividing the overall inference task into smaller, more manageable components.
Solution Approach 2:
The system dynamically adjusts the number and type of inference operations performed based on the confidence level of initial keyword matching and the context of the user input. When keyword matching is high-confidence, fewer inference operations are executed. When confidence is low or the input is ambiguous, additional inference operations are triggered. This adaptive approach balances inference accuracy with system complexity.
Data Source
AI summary
A vehicular voice recognition system for inferring an intention of a user includes: a storage storing an instruction use history, service use pattern information, and a service preferring item; a controller receiving an input instruction of the user and performing at least one of: a first inference operation of determining a service domain among a plurality of service domains corresponding to the input instruction and providing a service in the determined service domain, a second inference operation of providing a service based on the stored instruction use history, a third inference operation of providing a service based on the stored service use pattern information, and a fourth inference operation of providing a service based on the stored service preferring item; and an output unit provided in a vehicle outputting contents of the provided service using at least one of audio and images.