Offline Speech Recognition with Rejection Logic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech recognition systems, both online and offline, fail to provide effective results without a network connection or with unstable network conditions, and lack rejection capabilities.
Innovation Solution
A speech recognition method that involves obtaining features of a speech signal, performing a path search in a generated search space using a map established with developer-edited content, anti-models, and a lightweight language model, and determining the need for rejection based on decoding results to output either a rejection or recognition result.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If on-line speech recognition service is used, then recognition accuracy can be maintained, but the system cannot work without network or with unstable network
Solution Approach 1:
The speech recognition system is segmented into offline and online components. The offline speech recognition model is extracted and deployed locally on the terminal device, enabling independent operation without network dependency. This segmentation allows the system to maintain recognition functionality in both online and offline scenarios, resolving the contradiction between recognition accuracy and network availability.
2Adaptability or versatility
If conventional off-line speech recognition system is used, then network independence is achieved, but rejection effect is lost
Solution Approach 1:
The patent merges the offline speech recognition model with rejection processing capabilities into a unified local system. By integrating the acoustic model, language model, and rejection decision-making components on the terminal device, the system achieves both network independence and effective rejection functionality, simultaneously improving adaptability and maintaining precision.
3Reliability
If full speech recognition model is deployed offline, then rejection capability is achieved, but device complexity and resource consumption increase
Solution Approach 1:
The speech recognition model and rejection processing components are pre-configured and trained offline, then deployed as a packaged solution on the terminal device. This preliminary preparation allows the system to have ready-to-use rejection capabilities without requiring complex real-time training or configuration, reducing device complexity while maintaining reliable rejection functionality.
Data Source
AI summary
A speech recognition method and a speech recognition device are disclosed. The speech recognition method includes: obtaining features of a speech signal to be recognized; performing a path search in a search space generated by establishing a map according to the features to output a decoding result; judging whether a rejection is needed according to the decoding result; and when the rejection is needed, determining that a speech recognition result is the rejection, and when the rejection is not needed, obtaining the speech recognition result according to the decoding result. The method has a good recognition rejection effect.