Multi-Channel Speech Recognition Using Buffered Partial Results
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech interaction technologies are limited to processing single-channel speech signals, failing to meet the requirements of multi-user and personalized speech recognition, leading to low efficiency and increased interaction delay in multi-channel scenarios.
Innovation Solution
A method and apparatus that utilize a preset speech recognition model to process multiple channels of audio signals, leveraging stored recognition data from a buffer and generating partial recognition results, thereby reducing the need for full data processing by the model, enhancing efficiency and reducing resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the speech recognition model processes all audio signal channels from scratch, then recognition accuracy is maintained, but processing efficiency decreases and interaction delay increases
Solution Approach 1:
The patent performs preliminary speech recognition on the first audio signal channel and stores the recognition results in a buffer before receiving subsequent channels. When the second channel is received, the system retrieves previously stored recognition data instead of reprocessing, thereby reducing processing time and interaction delay while maintaining recognition accuracy.
2Adaptability or versatility
If the speech recognition model processes multiple channels of audio signals, then multi-user speech recognition capability is achieved, but resource consumption increases
Solution Approach 1:
The patent merges the processing of multiple audio signal channels by storing recognition results from the first channel in a shared buffer and retrieving them for the second channel. This combining approach allows the system to handle multi-user speech recognition scenarios while reducing redundant computational operations and lowering overall resource consumption.
3Productivity
If the system stores recognition data in a buffer for reuse, then processing efficiency improves, but memory usage increases
Solution Approach 1:
The patent extracts only the necessary recognition data from the speech recognition process and stores it in a buffer for future retrieval. By selectively storing only the essential recognition results rather than all processing data, the system improves processing efficiency for subsequent channels while keeping memory usage at manageable levels.
Data Source
AI summary
Disclosed are a speech interaction method, computer readable storage medium, and electronic device, improving processing efficiency in multi-channel speech interaction scenarios with low resource consumption and processing delay. The method includes: obtaining at least one channel of audio signal; recognizing the at least one channel of audio signal using a preset speech recognition model, to obtain a first type of recognition result; determining stored recognition data from a buffer; generating a second type of recognition result based on the stored recognition data; processing the first and second type of results using the model, to obtain at least one statement recognition result corresponding to the at least one channel of audio signal; performing semantic parsing on the statement recognition result, to obtain at least one parsing result; and generating, based on the at least one parsing result, an instruction for controlling a speech interaction device to perform a corresponding function.