Gesture recognition method based on interval distribution probability characteristics

Abstract

The invention relates to a gesture recognition method based on interval distribution probability characteristics. The method comprises the steps of S1, based on the infrared sensing technology, acquiring the infrared reflection signal and the infrared temperature signal of a gesture at a preset frequency f; S2, judging whether the gesture is a target action or not based on the infrared temperature variation; S3, automatically detecting the motion segment of the gesture, and conducting the denoising, normalizing, interpolating and pull-in treatment on the motion segment of the gesture to obtain a new data segment; S4, extracting features, and establishing a gesture feature template as a training sample based on the original data of a standard gesture at the same time; S5, finally, recognizing and testing the type of the sample gesture based on the KNN method. The above method is simple and intuitive, while the supervised learning process is avoided. The method is high in recognition rate for common dynamic gestures and good in real-time performance. Moreover, the method is large in difference inclusiveness during the varying process of environmental factors or gestures and is better in robustness.

Description

technical field [0001] The present invention relates to the technical field of human-computer interaction, and more specifically, relates to a gesture recognition method based on interval distribution probability features. Background technique [0002] Gesture recognition technology is currently a research hotspot in the field of human-computer interaction and has broad application prospects. Due to the polysemy diversity of gesture itself and the differences in time and space, gesture recognition technology still has a lot of room for development. Traditional gesture recognition mainly includes contact based on wearable sensors and non-contact based on computer vision. The recognition method based on wearable sensors requires users to wear professional equipment, is not affected by changes in the external environment, has a small amount of calculation, and has better real-time performance, but has disadvantages such as inconvenient use and high cost. Although the recogniti...

AI Technical Summary

Problems solved by technology

The recognition method based on wearable sensors requires users to wear professional equipment, is not affected by changes in the external environment background, has a small amount of calculation, and has better real-time performance, but has disadvantages such as inconvenient use and high cost.

Although the recognition method based on computer vision solves the problem of scope of application, it has strict requirements on equipment and environment, and the algorithm is more complicated and the real-time performance is poor.

[0003] At present, there are mainly recognition methods such as template matching, feature extraction, neural network and hidden Markov model for gesture recognition, which are widely adopted, but these recognition algorithms are relatively complex or require large storage space. With the increase of gesture recognition, the system storage space is not enough Or the recognition speed is too slow to affect the interaction efficiency, and it is difficult to meet the requirements of online recognition

Images