Dual Mode Speech Recognition System With Latency Cutoff
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ASR systems for mobile devices face challenges in balancing the responsiveness of local systems with the quality of server-based systems, often resulting in suboptimal speech recognition results due to limitations in processing and data storage, or delays associated with network dependencies.
Innovation Solution
A dual mode speech recognition method and system that utilizes both local and remote recognition systems. The system processes spoken queries through a local recognition system on the mobile device and a remote recognition system via a network connection, with a latency timer set to ensure timely results. If both systems recognize the query within the timer period, the result with the higher score is chosen; otherwise, the first recognized result is selected, and if necessary, the remote system updates the local vocabulary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If local recognition system is used, then speed and responsiveness are improved, but speech recognition quality deteriorates due to limited processing and data storage capabilities
Solution Approach 1:
The speech recognition system is segmented into two independent components: a local recognition system for rapid response and a remote recognition system for high-quality processing. Each component operates autonomously and can handle different types of recognition tasks based on their respective capabilities.
Solution Approach 2:
The patent combines the local and remote recognition systems into a unified dual-mode architecture that leverages the strengths of both approaches. The system dynamically selects or integrates results from either mode to achieve both speed and quality.
2Measurement precision
If remote recognition system is used, then speech recognition quality is improved, but response time deteriorates due to network communication delays
Solution Approach 1:
The system dynamically adjusts its operation mode based on real-time conditions such as network availability, task complexity, and quality requirements. It can switch between local-only mode, remote-only mode, or hybrid mode to optimize the balance between quality and response time.
Solution Approach 2:
The local recognition system performs preliminary speech recognition processing immediately upon receiving input, providing a fast initial response. The remote system then performs additional processing or verification if needed, rather than waiting for the local system to complete all processing.
3Speed
If dual mode recognition is implemented, then speech recognition quality and speed are both improved, but system complexity increases
Solution Approach 1:
A coordination module acts as an intermediary between the local and remote recognition systems, managing task distribution, result integration, and mode switching. This mediator simplifies the overall system architecture by providing a clear interface and control mechanism between the two recognition modes.
4Adaptability or versatility
If local vocabulary is updated with remote vocabulary, then adaptability is improved, but local storage requirements increase
Solution Approach 1:
Instead of downloading the complete remote vocabulary to the local device, the system selectively transfers only the specific words or phrases that are relevant to the user's needs and context. This partial action approach provides sufficient adaptability while minimizing local storage requirements.
Data Source
AI summary
A system and method is presented for performing dual mode speech recognition, employing a local recognition module on a mobile device and a remote recognition engine on a server device. The system accepts a spoken query from a user, and both the local recognition module and the remote recognition engine perform speech recognition operations on the query, returning a transcription and confidence score, subject to a latency cutoff time. If both sources successfully transcribe the query, then the system accepts the result having the higher confidence score. If only one source succeeds, then that result is accepted. In either case, if the remote recognition engine does succeed in transcribing the query, then a client vocabulary is updated if the remote system result includes information not present in the client vocabulary.