Speech Pronunciation Attribution for Shared Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computing devices with speech recognition functionality face challenges in accurately identifying individual users, especially in shared device scenarios, leading to undesirable interactions and reduced transcription accuracy.
Innovation Solution
The method involves determining candidate user profiles associated with a shared device by analyzing relationships such as login history, device proximity, and social connectivity, and then comparing pronunciation attributes to accurately match spoken utterances with the correct user profile, thereby providing appropriate audio responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the device compares pronunciation attributes against all possible user profiles, then user identification accuracy improves, but computational load and processing time increase
Solution Approach 1:
The patent segments the user identification process into two stages: first filtering candidate user profiles based on device association relationships (login history, device proximity, social connectivity), then comparing pronunciation attributes only against this reduced candidate set. This segmentation reduces the computational complexity from O(n) comparisons against all profiles to O(k) comparisons where k << n, while maintaining identification accuracy.
Solution Approach 2:
The patent performs preliminary filtering of candidate user profiles before the pronunciation attribute comparison step. By pre-establishing device association relationships and identifying candidate profiles in advance, the system prepares a reduced search space that accelerates the subsequent pronunciation matching process without sacrificing accuracy.
2Measurement precision
If the device processes and compares pronunciation attributes for all candidate user profiles, then transcription accuracy improves, but energy consumption increases
Solution Approach 1:
The patent divides the energy-intensive pronunciation comparison operation into two phases: an initial filtering phase that identifies candidate profiles using lightweight relationship data, and a subsequent comparison phase that processes pronunciation attributes only for candidates. This segmentation reduces total energy consumption by avoiding unnecessary processing of non-candidate profiles while maintaining transcription accuracy.
3Measurement precision
If the device maintains detailed relationship records for all user profiles, then user identification accuracy improves, but device complexity increases
Solution Approach 1:
The patent extracts only the necessary relationship indicators (login history, device proximity, social connectivity) from complete user profile data to identify candidate profiles. By taking out only these specific relationship attributes rather than processing entire profile structures, the system reduces device complexity while maintaining sufficient information for accurate user identification.
4Adaptability or versatility
If the device provides personalized audio responses for each user, then user experience quality improves, but processing time increases
Solution Approach 1:
The patent performs preliminary identification of candidate user profiles and their associated pronunciation attributes before generating audio responses. By preparing the candidate set and relevant pronunciation data in advance, the system enables rapid selection and personalized response generation, improving both user experience quality and response time.
Data Source
AI summary
Methods, systems, and apparatus for determining candidate user profiles as being associated with a shared device, and identifying, from the candidate user profiles, candidate pronunciation attributes associated with at least one of the candidate user profiles determined to be associated with the shared device. The methods, systems, and apparatus are also for receiving, at the shared device, a spoken utterance; determining a received pronunciation attribute based on received audio data corresponding to the spoken utterance; comparing the received pronunciation attribute to at least one of the candidate pronunciation attributes; and selecting a particular pronunciation attribute from the candidate pronunciation attributes based on a result of the comparison of the received pronunciation attribute to at least one of the candidate pronunciation attributes. With the methods, systems, and apparatus, the particular pronunciation attribute, selected from the candidate pronunciation attributes, is provided for outputting audio associated with the spoken utterance.