Context-Aware Voice Utterance Processing with Domain Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Electronic devices face limitations in processing user utterances due to hardware constraints, leading to delayed processing speeds and reduced accuracy in voice recognition services, with significant storage requirements and high costs for data management.
Innovation Solution
An electronic device equipped with a microphone, memory, and processor that acquires voice signals, determines domain sets based on context information, and performs operations using selected domain sets, allowing for efficient management and high-accuracy voice services by downloading user-selected domain sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the electronic device stores speech recognition service-related models, various resources, and databases locally, then the voice recognition accuracy is improved, but the storage space consumption and data download cost increase significantly
Solution Approach 1:
The patent segments the speech recognition service into multiple domain-specific models (e.g., weather domain, news domain, entertainment domain). Instead of storing all models locally, the device stores only the currently needed domain model and downloads other models on-demand. This segmentation allows the device to maintain high recognition accuracy for the active domain while minimizing storage space consumption.
Solution Approach 2:
The patent implements preliminary action by pre-loading or pre-caching frequently used domain models in the device's memory before they are actually needed. The system predicts which domains the user is likely to use based on historical data and user preferences, and prepares these models in advance. This reduces the need for real-time downloads and improves response speed while maintaining accurate voice recognition.
2Speed
If the electronic device processes only a limited number of user utterances locally, then the processing speed is improved, but the accuracy of voice recognition service deteriorates due to transmission delays to external servers
Solution Approach 1:
The patent divides speech recognition processing into two segments: simple, frequent utterances are processed locally by on-device models for fast response, while complex or uncertain utterances are transmitted to external servers for processing. This segmentation allows the system to maintain high processing speed for common commands while preserving accuracy for complex queries by leveraging server-side capabilities.
Solution Approach 2:
The patent introduces an intermediary mechanism where the device first attempts local processing using domain-specific models. When the local model's confidence threshold is not met or the utterance complexity exceeds local capabilities, the system acts as an intermediary by transmitting the utterance to external servers and relaying the results back to the user. This intermediary approach optimizes the balance between speed and accuracy.
3Adaptability or versatility
If the electronic device downloads and stores multiple domain sets for different user contexts, then the adaptability to different user intents is improved, but the memory management complexity increases
Solution Approach 1:
The patent implements dynamic memory management where the device automatically adjusts which domain models are stored in memory based on real-time user context, historical usage patterns, and available storage space. The system dynamically loads, unloads, and replaces domain models as needed, rather than maintaining a static set of all possible models. This dynamic approach enables high adaptability to different user intents while keeping memory management complexity manageable through automated decision-making algorithms.
Data Source
AI summary
An electronic device according to an embodiment of the invention includes: a microphone; a memory storing a plurality of domain sets; and at least one processor electrically connected to the microphone and the memory, wherein the at least one processor is configured to: acquire a voice signal using the microphone; acquire context information associated with at least one of the electronic device or a user; determine a first domain set of the plurality of domain sets based on at least the context information; and perform an operation corresponding to the voice signal based on the determined first domain set.