Collaborative man-machine hybrid intelligent recognition method

Abstract

The invention discloses a collaborative man-machine hybrid intelligent recognition method, which takes the process as a driving center, takes the operation as an information medium, enables machine recognition and artificial recognition to be riveted together in a real-time collaboration mode, combines speed, cost and reliability advantages of machine intelligent recognition and cognition, association, analysis and identification advantages of artificial recognition, and avoids defects of machine intelligent recognition in quality control and broad-spectrum applicability and the weakness of easy fatigue, easy manual mistake making and emotional fluctuation of artificial recognition at the same time. The collaborative man-machine hybrid intelligent recognition method specifically comprises the steps of 1) acquisition, 2) machine classification, 3) machine slicing, 4) machine pre-recognition, 5) border detection, 6) rule detection, 7) machine recognition, 8) sampling detection, 9) combination output, 10) artificial classification, 11) artificial label cutting, 12) artificial detection, 13) artificial recognition, and 14) artificial checking. The collaborative man-machine hybrid intelligent recognition method can be widely applied to voice and graphic text recognition.

Description

technical field [0001] The present invention relates to intelligent technology and IT technology, and specifically relates to a system that can cooperate and combine machine recognition ability and manual recognition ability in a systematic and standardized manner, so that intelligent recognition can simultaneously utilize the fast, low-cost, and reliable features of machine recognition. IT technologies and methods with high efficiency and wide-spectrum applicability in cognition, association, analysis, and identification with artificial recognition. Background technique [0002] The so-called intelligent recognition is to extract the characteristic content required in the input data according to the recognition requirements, according to certain patterns, models and methods, and convert the characteristic content into the target format for output. Generally, common intelligent recognition includes speech recognition (Speech To Text, which converts voice content into text), ...

AI Technical Summary

Problems solved by technology

[0003] People in the industry know that traditional intelligent recognition relies on a single machine intelligent recognition, because the principle obstacle of machine recognition is that the machine cannot autonomously recognize the validity of the recognition results, and machine intelligence lacks the ability of independent cognition and self-perception, and cannot control the quality of recognition , there will be significant quality fluctuations when encountering less training and learning edge data or application domains

For example, in speech recognition, when there are dialects or spoken expressions that have not been fully trained, the machine cannot recognize meaningful results; in graphic recognition, if the scanning or shooting angle of a form deviates greatly from the normal, the machine cannot Identify the correct text content

The limitations of machine intelligent recognition at this stage lead to the need for targeted technical optimization and improvement and training and learning of a large amount of data for any recognition application, which makes intelligent recognition unable to quickly enter new fields and solve recognition problems with broad spectrum application. Great restrictions restrict the application field, application depth and application range of intelligent identification

Images