Distributed Voice Recognition Using Client-Server Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing distributed speech recognition systems face challenges in maintaining recognition accuracy for both speaker-dependent and speaker-independent models, especially in noisy environments and when dealing with device-specific characteristics, leading to decreased performance.
Innovation Solution
The implementation of a distributed voice recognition system that coordinates speaker-specific and speaker-independent recognition using top-down and bottom-up processing principles, where the client-side processes noise suppression and device-specific optimization, and the server-side handles language processing with separate confidence scoring and a phonotactic table to validate results, effectively combining local and remote processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speaker-dependent recognition is used at the client device, then recognition accuracy for the specific user is improved, but general applicability to other speakers is lost
Solution Approach 1:
The system segments the recognition task between client and server: the client performs speaker-dependent acoustic recognition to capture user-specific characteristics, while the server performs speaker-independent language modeling to maintain general applicability. This segmentation allows both specialized and general capabilities to coexist.
Solution Approach 2:
The system merges results from both speaker-dependent client-side recognition and speaker-independent server-side recognition. By combining these complementary approaches, the system achieves both high accuracy for the specific user and robust general applicability across different speakers.
2Adaptability or versatility
If speaker-independent recognition is used on the server, then general applicability is maintained, but accuracy in noisy environments and device-specific conditions decreases
Solution Approach 1:
The system divides the recognition pipeline so that the server handles speaker-independent language modeling for general applicability, while the client handles speaker-dependent acoustic recognition adapted to specific device characteristics and noise conditions, allowing each component to optimize for its strength.
Solution Approach 2:
The client device acts as an intermediary that adapts the general speaker-independent recognition from the server by applying speaker-specific and device-specific acoustic models, thereby bridging the gap between general applicability and specialized accuracy.
3Measurement precision
If distributed processing is implemented between client and server, then recognition accuracy is improved through specialized modeling, but system complexity increases
Solution Approach 1:
The system segments processing tasks between client and server based on their respective capabilities: acoustic recognition with speaker adaptation at the client, and language modeling at the server. This segmentation improves accuracy while distributing complexity appropriately across the distributed system.
Data Source
AI summary
A distributed voice recognition system (500) and method employs principles of bottom-up (i.e., raw input) and top-down (i.e., prediction based on past experience) processing to perform client-side and server-side processing by (i) at the client-side, replacing application data by a phonotactic table (504); (ii) at the server-side, tracking separate confidence scores for matches against an acoustic model and comparison to a grammar; and (iii) at the server-side using a contention resolver (514) to weight the client-side and server-side results to establish a single output which represents the collaboration between client-side processing and server-side processing.