With the AI ​​artificial intelligence AlphaGo and Li Shishi's five-man man-machine battle came to an end, the Alpha dog finally scored a 4:1 big score, Li Shishi, and people's curiosity about the Alpha dog was pushed to a climax. Alpha Dog also showed the power of deep learning neural network, DNN, in front of people.
So what is DNN? DNN is a hot topic in the field of artificial intelligence in recent years, and its applications in speech recognition, automatic driving, handwriting recognition, etc. are extremely successful. In the field of artificial intelligence, a neural network is a model that simulates a biological nervous system. It consists of a number of unidirectionally connected neurons that convert an input signal into an output signal based on various parameters of the linked neuron. Compared to ordinary neural networks, deep neural networks (DNN) use implicit multi-layer complex structures, as well as nonlinear transformations, to express a high degree of abstraction of data. These features are closer to the human brain and make it easier to implement certain functions of the human brain, such as the brain functions required for Go.
Alpha Dog uses two deep learning neural networks, namely “strategic network†and “value network†to fit the situational strategy function and evaluation function. It can be said that this is the two brains of Alpha Dog. The victory of Alpha Dogs detonated the concept of DNN. In today's big data, deep learning neural networks have been applied to a large number of fields, becoming an important tool for revealing scientific principles and upgrading existing industry business models.
So can such a cool DNN be applied to the hot VR industry? The answer is yes. At present, the application of DNN technology to VR field includes speech recognition, gesture recognition, etc. The author mainly introduces the DNN in the gesture recognition algorithm.
There are two main methods of gesture recognition: data glove-based gesture recognition method and computer vision-based gesture recognition method.
The computer vision-based gesture recognition method has become a hot spot in today's research because it does not depend on equipment, more natural human-computer interaction effects, and better immersion. Based on the computer vision gesture recognition, the gesture image information is obtained from the camera, and after appropriate data pre-processing, the gesture is segmented from the image, and the gestures of the segmented gesture are extracted, and the gesture template is used to classify.
The traditional classification methods include template matching, fingertip detection methods, etc., but these methods need to manually extract the target features from the image and write templates to match. With the deep learning neural network being proposed, gesture recognition based on deep learning neural network (DNN) has emerged. This gesture recognition algorithm has great advantages in all aspects compared with traditional gesture recognition algorithms.
The deep learning neural network works from the analogy of the brain mechanism of the same person. This process from the original input to the higher level of abstract iteration gives the model a high degree of abstraction, making the deep learning neural network very effective. Extracting the feature information of the data from a large amount of tagged data, fully exploiting the intrinsic properties of the data and valuable characterization data, and then combining the lower-level features into more abstract high-level features, while the advanced features are more advanced and more essential data. Description, so that better results can be obtained on the classification problem.
As we all know, a major technical problem faced by the VR industry is the processing of massive data. This is especially true for gesture recognition technology. There are many joints in the hand, which require a very strong recognition ability to accurately identify each fine action. The multi-hidden layer structure of the deep learning model enables the model to effectively use massive data for training. The more data used, the higher the model performance, which is very suitable for gesture recognition in VR environment.
In computer vision-based gesture recognition, traditional algorithms cannot directly and effectively extract information useful for the target from the image. The learning ability of deep learning is extremely powerful, and even the complex low-resolution images can extract the target depth features well. The image background required for gesture recognition based on DNN does not need to be fixed, and the algorithm allows the existence of a moving background even within a certain range, thereby improving the environmental tolerance and fineness of recognition.
| How is DNN implemented in gesture recognition?Next, the author will introduce the specific implementation method.
The traditional identification system process can be divided into three steps: detection, identification, and tracking. The specific implementation process is: sensor acquisition information, preprocessing, feature extraction, feature selection, and finally final reasoning, prediction, or identification. It is generally believed that the last part is part of machine learning. This part is the essence of the whole system. Whether it can learn effective knowledge from the data is directly related to whether the whole system can work as expected, but also has the characteristics of the early stage. Extract or select bad conditions, which will affect system performance.
DNN-based gesture recognition is to let the machine extract features by itself, without manual feature extraction. The powerful learning ability makes the model achieve ideal results in complex backgrounds. The gesture recognition process is as follows:
Firstly, a gesture recognition acquisition system is created, and the moving target detection is performed, and part of the foreground, that is, the moving target is extracted while the detection is performed, and the gesture image is obtained. After the gesture image is collected, an appropriate network protocol is selected according to different application scenarios, and the data is transmitted to the computing platform to perform complex background gesture recognition using the deep learning algorithm, and finally the gesture recognition result is given.
What is the VR based on DNN?A VR product based on the idea of ​​DNN algorithm and using a single-machine GPU method to accelerate the training and recognition of deep networks. The gesture recognition module can realize the close-range 3D imaging of the hand using the depth camera. In combination with DNN, a set of data processing algorithms is developed independently, which realizes high-precision real-time hand motion recognition, which can track a single finger and recognize the subtle movement of each finger. It can also be extended to track multiple hands.
The gesture recognition of such a VR product has high capture precision, fast response speed, high sensitivity, and can be protected from ambient light, and can be used indoors and outdoors.
However, there are currently no DNN-based gesture recognition products on the market. The StepVR product independently developed by G-Wearables, which I know, is based on the DNN algorithm, but I haven't seen it yet. I can look forward to it.
A business tablet must have powerful business functions. At present, users use tablet products more for entertainment, and insufficient development of commercial value has become a major factor restricting the development of tablet computers. Many far-sighted terminal manufacturers have set their sights on the business tablet field.
Generally speaking, when business users choose tablet PC products, they still focus on "regular items" such as processor, battery, operating system, and built-in applications, especially office software applications, which are a good helper for business users.
First of all, there must be powerful office software such as customer management software, which can record customer data and related information anytime, anywhere. There is also a need for powerful marketing software, software marketing, reducing manual intervention, and one person can handle all the publicity and customer information.
Business Tablet,4k Tablet,5G Tablet
Jingjiang Gisen Technology Co.,Ltd , https://www.jsgisengroup.com