Combined decoding method and system used for voice recognition systems
A speech recognition and joint decoding technology, which is applied in speech recognition, speech analysis, instruments, etc., can solve problems such as time-consuming, fusion effect cannot be guaranteed to be stable, sub-sequences are not aligned, etc., to achieve the effect of ensuring the fusion effect
Active Publication Date: 2019-04-02
AISPEECH CO LTD
View PDF7 Cites 10 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0008] ROVER, Confusion network combination, and minimum Bayes risk latticecombination methods all require multiple decodings. To be precise, it takes as many times as there are several systems, which is very time-consuming; although the joint decoding method only requires one decoding, for the CTC system, the system’s Posterior score sequences are often not aligned, and the fusion effect cannot be guaranteed to be stable
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0032] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.
[0033] It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.
[0034] The invention may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, progr...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a combined decoding method and a combined decoding system used for voice recognition systems, used for combined decoding between a first voice recognition system and a second voice recognition system. The method comprises the following steps: acquiring a first posterior score sequence of the first voice recognition system and a second posterior score sequence of the secondvoice recognition system; determining the aligning relation between the first posterior score sequence and the second posterior score sequence by adopting the DTW algorithm; fusing the first posteriorscore sequence and the second posterior score sequence according to the aligning relation, thus obtaining a fusing result; and carrying out decoding processing on the fusing result. According to themethod and the system provided by the invention, only one time of decoding is needed when the fusing of the voice recognition systems is carried out, and meanwhile, the aligning of posterior scores of the two system is realized before the results of the two systems are fused, so that the fusing effect is guaranteed.
Description
technical field [0001] The invention relates to the technical field of speech recognition, in particular to a joint decoding method and system for a speech recognition system. Background technique [0002] Connectionist temporal classification (CTC, connectionist temporal classification) has been successfully used in speech recognition tasks. Without frame-level annotations, CTC can automatically learn the correspondence between speech frames and phoneme annotations. Although this is beneficial to shorten the training process, it becomes difficult for the system to integrate at the frame level. [0003] The system fusion technologies for speech recognition mainly include ROVER, confusion network combination, minimum Bayes risk lattice combination, and joint decoding. (joint decoding). [0004] ROVER: After aligning the recognition results of each recognition system according to the minimum edit distance, select the word with the highest confidence in each slot; [0005] ...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G10L15/22G10L15/06G10L25/30
CPCG10L15/06G10L15/063G10L15/22G10L25/30
Inventor 俞凯钱彦旻郭嘉祺游永彬
Owner AISPEECH CO LTD