Distributed Speech Recognition Parallel Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition systems, particularly in distributed modes, face inefficiencies due to sequential processing at user terminals and servers, leading to prolonged recognition times and increased user request response times.
Innovation Solution
A distributed speech recognition method where both user terminals and servers process audio signals in parallel, with the terminal calculating and transmitting modeling parameters or audio signals, allowing the server to determine associated forms independently, thereby reducing overall recognition time by combining processing capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sequential processing is used at user terminals and servers, then processing reliability is improved, but recognition time increases
Solution Approach 1:
The terminal performs preliminary acoustic analysis to extract modeling parameters from the audio signal before transmission. This preliminary processing reduces the data size for transmission and allows the server to focus on recognition tasks, thereby reducing overall recognition time while maintaining processing reliability through distributed sequential processing.
Solution Approach 2:
The speech recognition system is segmented into two independent parts: acoustic analysis at the terminal and recognition at the server. This segmentation allows parallel execution of different processing stages, reducing total recognition time while maintaining the reliability of each segment through specialized processing.
2Productivity
If distributed processing is implemented, then processing speed is improved, but system complexity increases
Solution Approach 1:
Modeling parameters serve as an intermediary between the audio signal at the terminal and the recognition models at the server. This intermediary representation simplifies the distributed processing by providing a standardized data format that reduces transmission complexity and enables efficient recognition processing at the server.
3Loss of energy
If acoustic analysis is performed at the terminal, then transmission efficiency is improved, but terminal processing requirements increase
Solution Approach 1:
The audio signal is transformed from raw waveform data into compact modeling parameters through acoustic analysis at the terminal. This parameter transformation significantly reduces the amount of data that needs to be transmitted to the server, improving transmission efficiency while the terminal processing energy is offset by the substantial savings in network transmission energy.
Data Source
AI summary
This invention relates to a distributed speech recognition method comprising at least one user terminal and at least one server which can communicate with each other by means of a telecommunication network. The inventive method comprises the following steps consisting in: at the user terminal, attempting to associate a saved form with the signal to be recognized and, independently of said step, transmitting a signal to the server, indicating the signal to be recognized; and, at the server, attempting to associate a saved form with the signal received.