Microsoft Technology Licensing, LLC

Cuiling Lan

Jizheng Xu

Wenjun Zeng

Junliang Xing

Feng Wu

Chong Luo

Wentao Zhu

Wenxuan Xie

Sijie Song

Yanghao Li

Techniques for selectively transforming one or more coding units when coding video content are described herein. The techniques may include determining whether or not to transform a particular coding unit. The determination may be based on a difference in pixel values of the particular coding unit and/or one or more predefined rate-distortion constraints. When it is determined to not perform a transform, the particular coding unit may be coded without transforming the particular coding unit.

Non-transform coding

Various implementations of the subject matter described herein relate to a neural network-based action detection. There is provided an action detection scheme using a neural network. The action detection scheme can design and optimize the neural network model based on respective importance of different frames such that frames that are more important or discriminative for action recognition tend to be assigned with higher weights and frames that are less important or discriminative for action recognition tend to be assigned with lower weights.

Neural network-based action detection

In accordance with implementations of the subject matter described herein, a hybrid digital-analog coding scheme is proposed. In general, in accordance with implementations of the subject matter described herein, digital or analog encoding is selected at a level of chunks of a frame rather than at a level of the whole frame. Further, the encoding of a chunk is based on an expected distortion to be caused by analog transmission of the chunk over a communication channel, the distortion being estimated based on a constraint on available transmission resources over the transmission channel.

Hybrid digital-analog coding

In implementations of the subject matter described herein, an action detection scheme using a recurrent neural network (RNN) is proposed. Representation information of an incoming frame of a video and a predefined action label for the frame are obtained to train a learning network including RNN elements and a classification element. The representation information represents an observed entity in the frame. Specifically, parameters for the RNN elements are determined based on the representation information and the predefined action label. With the determined parameters, the RNN elements are caused to extract features for the frame based on the representation information and features for a preceding frame. Parameters for the classification element are determined based on the extracted features and the predefined action label. The classification element with the determined parameters generates a probability of the frame being associated with the predefined action label. The parameters for the RNN elements are updated according to the probability.

On-line action detection using recurrent neural network

In implementations of the subject matter described herein, a current captured frame of a video is compared with a respective reference frame to determine a correlation therebetween. The correlation is used to indicate a change degree of the current frame. If the correlation for the current frame is below a predetermined threshold, the current frame may not be directly determined as including a motion of an object. Instead, correlations between one or more frames before or after the current frame and their respective reference frames are took into account. If the correlations of the frames under consideration are below the predetermined threshold, it may be detected that the current frame includes a motion of an object. In this way, incorrect detection of the object motion is reduced in the cases when larger changes in frames of a video are caused by factors such as noise and error, and the accuracy of the object motion detection is improved.

Motion detection of object

In implementations of the subject matter described herein, an action detection scheme using a recurrent neural network (RNN) is proposed. Joint locations for a skeleton representation of an observed entity in a frame of a video and a predefined action label for the frame are obtained to train a learning network including RNN elements and a classification element. Specifically, first weights for mapping the joint locations to a first feature for the frame generated by a first RNN element in a learning network and second weights for mapping the joint locations to a second feature for the frame generated by a second RNN element in the learning network are determined based on the joint locations and the predefined action label. The first and second weights are determined by increasing a first correlation between the first feature and a first subset of the joint locations and a second correlation between the second feature and the first subset of the joint locations. Based on the joint locations and the predefined action label, a parameter for a classification element included in the learning network is also determined by increasing a probability of the frame being associated with the predefined action label. The probability is generated by the classification element at least based on the first and second features.

Growing community of inventors

Xi'an, China

Cuiling Lan