New hand gesture recognition algorithm strikes a stability between complexity, accuracy, and applicability

Within the 2002 science fiction blockbuster movie Minority Report, Tom Cruise’s character John Anderton makes use of his arms, sheathed in particular gloves, to interface along with his wall-sized clear laptop display screen. The pc acknowledges his gestures to enlarge, zoom in, and swipe away.

Though this futuristic imaginative and prescient for computer-human interplay is now 20 years outdated, in the present day’s people nonetheless interface with computer systems by utilizing a mouse, keyboard, distant management, or small contact display screen. Nonetheless, a lot effort has been devoted by researchers to unlock extra pure types of communication with out requiring contact between the consumer and the system. Voice instructions are a distinguished instance which have discovered their approach into trendy smartphones and digital assistants, letting us work together and management units by means of speech.

Hand gestures represent one other necessary mode of human communication that might be adopted for human–laptop interactions. Latest progress in digital camera programs, picture evaluation, and machine studying have made optical-based gesture recognition a extra enticing choice in most contexts than approaches counting on wearable sensors or knowledge gloves, as utilized by Anderton in Minority Report.

Nonetheless, present strategies are hindered by quite a lot of limitations, together with excessive computational complexity, low pace, poor accuracy, or a low variety of recognizable gestures. To sort out these points, a group led by Zhiyi Yu of Solar Yat-sen College, China, not too long ago developed a brand new hand gesture recognition algorithm that strikes a very good stability between complexity, accuracy, and applicability. As detailed of their paper, which was revealed within the Journal of Digital Imaging, the group adopted modern methods to beat key challenges and understand an algorithm that may be simply utilized in consumer-level units.

One of many important options of the algorithm is adaptability to completely different hand sorts. The algorithm first tries to categorise the hand sort of the consumer as both slim, regular, or broad primarily based on three measurements accounting for relationships between palm width, palm size, and finger size. If this classification is profitable, subsequent steps within the hand gesture recognition course of solely evaluate the enter gesture with saved samples of the identical hand sort.

Conventional easy algorithms are inclined to undergo from low recognition charges as a result of they can’t address completely different hand sorts. By first classifying the enter gesture by hand sort after which utilizing pattern libraries that match this kind, we are able to enhance the general recognition price with virtually negligible useful resource consumption.”

Zhiyi Yu, Solar Yat-sen College

One other key facet of the group’s technique is using a “shortcut function” to carry out a prerecognition step. Whereas the popularity algorithm is able to figuring out an enter gesture out of 9 potential gestures, evaluating all of the options of the enter gesture with these of the saved samples for all potential gestures could be very time consuming.

To resolve this drawback, the prerecognition step calculates a ratio of the world of the hand to pick the three most probably gestures of the potential 9. This easy function is sufficient to slim down the variety of candidate gestures to 3, out of which the ultimate gesture is set utilizing a way more complicated and high-precision function extraction primarily based on “Hu invariant moments.” Yu says, “The gesture prerecognition step not solely reduces the variety of calculations and {hardware} sources required but additionally improves recognition pace with out compromising accuracy.”

The group examined their algorithm each in a business PC processor and an FPGA platform utilizing an USB digital camera. They’d 40 volunteers make the 9 hand gestures a number of occasions to construct up the pattern library, and one other 40 volunteers to find out the accuracy of the system. General, the outcomes confirmed that the proposed method might acknowledge hand gestures in actual time with an accuracy exceeding 93%, even when the enter gesture pictures had been rotated, translated, or scaled. In line with the researchers, future work will deal with bettering the efficiency of the algorithm below poor lightning situations and rising the variety of potential gestures.

Gesture recognition has many promising fields of software and will pave the way in which to new methods of controlling digital units. A revolution in human-computer interplay is perhaps shut at hand!


SPIE–Worldwide Society for Optics and Photonics

Journal reference:

Zhang, Q., et al. (2021) Hand gesture recognition algorithm combining hand-type adaptive algorithm and effective-area ratio for environment friendly edge computing. Journal of Digital Imaging.

#hand #gesture #recognition #algorithm #strikes #stability #complexity #accuracy #applicability