Distributed Speech Recognition Synchronization via Readiness Notification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed Speech Recognition (DSR) systems face challenges in echo prevention and bandwidth conservation due to varying network delays and limited bandwidth, especially in mobile channels, which degrade speech recognition accuracy and user experience.
Innovation Solution
The system addresses these issues by having a DSR client receive notification from a voice server of readiness to conduct speech recognition, allowing it to transmit speech only after confirmation, thereby reducing echo effects and conserving bandwidth by using Voice Activity Detection (VAD) and discontinuous transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If speech data is transmitted continuously over voice channels, then speech recognition can be performed, but network bandwidth is wasted and echo effects occur due to varying delays
Solution Approach 1:
The system uses Voice Activity Detection (VAD) to detect periodic speech segments and transmits speech data only during these periodic intervals when speech is actually present, rather than continuous transmission. This periodic action based on speech activity patterns reduces bandwidth consumption while maintaining recognition accuracy.
Solution Approach 2:
The DSR client autonomously determines when to transmit speech data by performing local Voice Activity Detection and responding to server readiness notifications without requiring continuous server polling or manual control. This self-service approach optimizes transmission timing based on local speech detection, reducing unnecessary bandwidth usage.
2Speed
If speech recognition starts immediately upon receiving speech input, then response time is reduced, but echo effects from network delay degrade recognition accuracy
Solution Approach 1:
The voice server sends a readiness notification to the DSR client before the client begins transmitting speech data. This preliminary action allows the server to prepare recognition resources in advance, and the client to synchronize transmission start time, eliminating echo effects while maintaining fast response through pre-coordinated timing.
Solution Approach 2:
The server provides feedback in the form of a readiness notification to the client, creating a feedback loop that synchronizes the start of speech transmission. This feedback mechanism ensures the server is prepared to receive and process speech data immediately when transmitted, preventing echo effects while maintaining rapid response.
3Reliability
If echo removal is performed on the DSR client, then recognition accuracy improves, but perfect echo removal is difficult to achieve due to stochastic packet switching delays
Solution Approach 1:
Instead of attempting complex echo removal after speech transmission, the system performs preliminary synchronization by having the server send a readiness notification before the client transmits speech data. This preliminary coordination prevents echo effects at their source, avoiding the need for complex post-processing echo removal algorithms.
Solution Approach 2:
The patent converts the potential harm of network delay into a benefit by using the server's readiness notification as a synchronization signal. The delay inherent in packet switching is transformed into a coordinated timing mechanism where the server prepares in advance, and the client transmits precisely when the server is ready, turning the delay problem into a synchronization solution.
4Loss of energy
If push to talk switching is used to conserve bandwidth, then bandwidth consumption is reduced, but user experience deteriorates due to manual switching requirements
Solution Approach 1:
The system employs automatic Voice Activity Detection on the client side to determine when speech is present and transmission should occur. This self-service approach eliminates the need for manual push-to-talk switching, maintaining natural user interaction while achieving bandwidth conservation through intelligent, automatic detection of speech segments.
Data Source
AI summary
Methods, apparatus, and computer program products are disclosed for synchronizing distributed speech recognition (‘DSR’) that include receiving in a DSR client notification from a voice server of readiness to conduct speech recognition and, responsive to the receiving, transmitting by the DSR client, from the DSR client to the voice server, speech for recognition.