WiMi Hologram Cloud Inc (WIMI)

[StockpickerCAPC]

WIMI (NASDAQ: WIMI) develops a virtual digital human gesture generation algorithm based on application scenarios.

Source
https://cj.sina.com.cn/articles/view/7651844612/1c815e20402001g024?from=finance

March 10, 2023

With the blessing of virtual technology, big data, artificial intelligence and other technologies, virtual digital humans are constantly evolving, with increasingly realistic images, wider application ranges, and greater commercial value. "Digital Human In-depth Industry Report" predicts that by 2030, the overall market size of my country's virtual digital human will reach 270 billion yuan.

A virtual digital human is a composite of human characteristics such as appearance, performance, and interaction, based on technologies such as computer graphics, graphics rendering, motion capture, deep learning, and speech synthesis. It has human geometric and behavioral characteristics . From production to digitization of appearance, virtual digital human gradually deepens into interactive behavior and intellectualization of thought. It not only has human appearance and behavior, but also has human thought, can recognize the external environment, and interact naturally with people.

Gesture, as an important interaction method, has been widely used in car navigation, virtual simulation and other fields. As a new generation of interactive platform, virtual digital human can integrate other interactive technologies to provide multi-modal interactive experience. By establishing a mapping between the abstract communication intention and the physical realization of gestures, and generating rich gestures, the emotional expression of virtual humans can be enriched.

It is understood that WIMI Hologram (NASDAQ: WIMI) is developing a virtual digital human gesture generation algorithm system based on application scenarios, which allows virtual digital human to produce different movements through changes in the environment. Not every sentence of the virtual digital human must be accompanied by gestures, and the gestures of the same sentence in different scenarios will be different. Therefore, the application scenarios of the virtual digital human need to be deeply designed.

The abstract communication intention is processed by natural language, and then the text is mapped to the classification of gesture semantics to build a semantic classification model. Firstly, it is judged whether gestures are needed, because gestures are used to assist and enhance semantic expression, and when it is uncertain whether gestures are needed, it tends to give the result that gestures are not needed. Then make statistics on the corpus of different scenes constructed, analyze the patterns of all actions in the corpus and their corresponding relationship with semantics, sort out the semantic actions in the corpus, and carry out emotional, gesture metaphor semantics, original text and gesture quantitative description language The mapping relationship, and build a classification model.

The mapping process from semantics to quantitative descriptions of gestures is a one-to-many classification problem. For different application scenarios, different corpora are used for training. For multiple gestures under the same metaphorical gesture semantic subdivision category, natural language processing technology is used to semantically match the gesture text with the input text, and the best matching gesture is selected. For the purpose of communication, through natural language understanding, the classification of the text to emotion and gesture metaphor semantics, and the association relationship with the original text are constructed, and finally a virtual digital human gesture is generated.

Gesture interaction of virtual digital human greatly enhances its emotional expression. Based on the automatic generation algorithm of virtual digital human gestures based on application scenarios, WIMI uses the semantic classification method of metaphorical gestures to construct a quantitative description language for gestures, provides a quantitative method for the computable semantics of gestures, and proposes different application scenarios. The construction method of emotional corpus and the construction of emotional corpus provide a data basis for the research of virtual human gesture generation algorithm.

Generally speaking, the theory and technology of virtual digital human are becoming more and more mature, and the scope of application is also expanding. Virtual digital humans have been applied in many industries such as finance, transportation, logistics, retail, manufacturing, etc., helping different industries realize digital transformation.