Intention-Based Voice Recognition on Mobile Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional voice dialogue systems on mobile communication devices with limited CPU and memory resources restrict users to a limited set of prescribed commands, making it difficult for them to communicate their intentions effectively, especially when they cannot remember exact sentences or use their usual vocabulary.
Innovation Solution
A method and apparatus for intention-based communications that includes receiving user input, converting it into linguistic representations, generating a phoneme lattice, scoring stored intention n-grams against the lattice, determining the highest scoring intention, and executing it if above a threshold, allowing users more flexibility in choosing words to communicate their intentions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a voice dialogue system is implemented on the server side, then flexibility is improved allowing users to choose words, but device complexity and resource consumption increase
Solution Approach 1:
The system divides the voice processing workload between the mobile device and the server. The device performs local phoneme lattice generation and intention scoring, while the server handles more complex language model processing. This segmentation allows flexibility to be distributed, improving user freedom to choose words without concentrating all complexity in one location.
Solution Approach 2:
The patent introduces an intermediary intention recognition layer that translates user speech into high-level intentions rather than requiring exact command matching. This intermediary layer acts as a buffer between the user's flexible language choices and the system's structured command execution, enabling natural language processing without requiring the entire system to be complex.
2Use of energy by moving object
If recognition grammars are restricted to limited prescribed commands, then device resource consumption is reduced, but ease of operation deteriorates as users must follow specific commands
Solution Approach 1:
The system performs partial intention recognition locally on the device using a reduced set of operations (phoneme lattice generation and intention scoring), while leaving more comprehensive language understanding to the server. This partial action approach allows the device to conserve resources while still providing improved ease of operation compared to fully restricted grammars.
Solution Approach 2:
The patent changes the parameters of local processing by using phoneme-level analysis instead of full word-level recognition, and by scoring pre-defined intention patterns rather than performing complete natural language understanding. This parameter change reduces computational complexity and power consumption while maintaining improved ease of use through more flexible input acceptance.
3Ease of operation
If full natural language processing is implemented, then ease of operation is improved, but device resource consumption increases beyond available CPU and memory
Solution Approach 1:
The system segments the natural language processing pipeline into device-side components (phoneme lattice generation, intention scoring) and server-side components (comprehensive language understanding). This segmentation enables the device to provide improved ease of operation through flexible input acceptance while consuming manageable CPU resources by offloading intensive processing to the server.
Data Source
AI summary
A method and apparatus for intention based communications in a mobile communication device is disclosed. The method may include receiving an input from a user of the mobile communication device, converting speech portions in the user's input into linguistic representations, generating a phoneme lattice based on the linguistic representations, scoring stored intention n-grams against the generated phoneme lattice, scoring intentions from the intention n grams, determining the highest scoring intention, determining whether the highest scoring intention is above a predetermined threshold, wherein if the highest scoring intention is above the predetermined threshold, executing the determined intention.