34 results about "Spatiotemporal encoding" patented technology

The invention discloses a fight recognition method and device based on a dual-channel cross-attention mechanism. The method first collects and generates video data sets, sets two frame rates of fast and slow to obtain different frame image sequences, and sends them to the fast and slow channels respectively after preprocessing. The fast and slow channel adopts the Transformer encoder based on the separate spatiotemporal self-attention mechanism to extract the spatiotemporal encoding features of the image sequence; then, through the cross-attention module, the CLS token of one channel and the patch token information of the other channel are fused to realize the fusion of dual-channel spatiotemporal encoding features; finally , the fused spatio-temporal coding features are passed through the multi-layer perception machine head for fighting behavior recognition. The invention can effectively extract the spatio-temporal features of the video through the dual-channel Transformer model and the cross-attention module, improve the accuracy of fighting behavior recognition, and is suitable for indoor and outdoor monitoring systems.

The invention discloses a single scan space-time coding imaging reconstruction method based on the residual network, which relates to a magnetic resonance imaging reconstruction technique based on a deep learning network. The invention provides a single scan space-time coding imaging reconstruction method based on a residual network for obtaining a two-dimensional image in a single scan and reconstructing by a deep learning method. The excitation pulse is replaced by a linear sweep pulse, which effectively resists image distortion caused by the nonuniform magnetic field and chemical shift, andat the same time obtains imaging speed, resolution and signal-to-noise ratio similar to EPI. The SPEN imaging is undersampled along the phase encoding direction. Although the space-time coding imaging signal itself can reflect the contour of the imaged object without reconstruction, the inherent resolution of the contour is typically very low. The single scan space-time coding imaging reconstruction method based on residual network utilizes deep learning to reconstruct SPEN images from low-resolution signal space, greatly improves image resolution, exhibits proton density distribution, and obtains high signal-to-noise ratio while obtaining a resolution similar to conventional deconvolution reconstruction methods.

The invention provides a water-fat separation and rebuilding method based on single-scanning space-time coding magnetic resonance imaging, and relates to a magnetic resonance imaging method. The method comprises the steps of utilizing a single-scanning space-time coding and imaging sequence to encode and collect a water-fat signal, wherein a 90-degree linear frequency sweep pulse is combined with a space-time coding gradient to give one quadratic phase to the water-fat signal in an exciting stage; based on the characteristic, obtaining priori knowledge images of water and fat used for guiding water-fat separation and rebuilding by selecting an appropriate filter to preprocess collected data; in a water-fat separation stage, utilizing the priori knowledge images as an initial value and weighted information, and obtaining a high-resolution and high-signal-to-noise ratio water image and fat image by solving a super-resolution lifting and edge artifact removing algorithm equation in a water-fat environment. According to the water-fat separation and rebuilding method based on single-scanning space-time coding magnetic resonance imaging, the water-fat separation of single-scanning space-time coding magnetic resonance imaging can be effectively achieved, meanwhile the image quality of separated images is improved, and the clinical application progress is promoted.

A disclosed transmitter for wireless communication includes multiple transmitting antennas, a symbol mapper for mapping an input block including multiple binary bits and representing information to be transmitted to a symbol representing an ordered plurality of complex numbers, a space-time encoder for applying an encoding operator to the symbol to produce a vectorized space-time codeword defining electrical signals to be transmitted by the transmitter, the encoding operator being dependent on a set of predefined stabilizer generators, and circuitry to collectively transmit, by the antennas to multiple receiving antennas of a receiver over a wireless transmission channel, the electrical signals defined by the vectorized space-time codeword. The receiver includes a space-time decoder for recovering the symbol from the electrical signals transmitted by the transmitter using a decoding operation that is based on maximum likelihood inference, and a symbol de-mapper for recovering the input block from the symbol.

The radio transmission device of the present invention is provided with: a plurality of transmission antennas; The initial symbol generating part (2), as the reference signal when differential encoding starts, generates the initial symbol; The differential encoding unit (differential encoding section 3, transmission signal power Calculation section 6, one-square-root calculation section 7), carry out differential encoding to the transmitted symbol mapped with information bit to generate the symbol after differentiation; Perform space-time encoding to generate signals transmitted from each transmitting antenna, and the differential encoding unit determines the differential encoding generated in this differential encoding process based on the power of the initial symbol or the power of the differentially differentiated symbol generated last time. The power of the following symbol.

Described here are systems and methods for volumetric excitation in magnetic resonance imaging (“MRI”) using frequency modulated radio frequency (“RF”) pulses. In general, quadratic phase modulation along the slice encoding direction is implemented for additional spatiotemporal encoding, which better distributes signal content in the slice direction and enables higher acceleration rates that are robust to slice-undersampling.

The invention discloses a transceiving method and a realization device thereof used for single frequency network signals in the technical field of digital television networks, which comprise the steps as follows: a PN frame header after PN frame header time coding processing and transmitting signal combined processing is inserted in front of a data section at a transmitting end; moreover, relevant PN processing and channel estimation processing are implemented at a receiving end. Aiming at solving a problem that received signals can not be effectively resolved due to multipath interferences caused by the single frequency network in overlapping areas, coding processing is implemented on the PN frame header before modulation, and then data jointing and demodulating transmitting are implemented; during receiving, a transmitting signal is obtained through relevant processing and channel estimation on the received and coded PN frame header as well as known PN frame header. The invention greatly improves the signal-to-noise of the channels, and can obtain an effective signal from the interfered signals more easily.

The invention discloses a video prediction method based on a time-space propagation hierarchical codec. The method of the present invention first samples the given original video to obtain the frame sequence and inputs it to the low-level visual memory encoder to obtain the low-level visual coding features and low-level memory state features; features, and extract high-level semantic features through a high-level semantic encoder; then fuse the obtained low-level visual coding features, low-level memory state features, and high-level semantic coding features through a hierarchical feature decoder to output predicted video frames. The method of the present invention can not only integrate the low-level visual memory features and high-level semantic features, but also make the low-level visual information spread along the time sequence direction through the spatio-temporal propagation module, and can also use the prior knowledge of the first frame of the video to solve the video frame blur problem to a certain extent, Improve overall clarity and visual quality of predicted video frames.

One or more embodiments of this specification disclose a space-time coding method, a space-time index and query method and device, which are used to realize efficient organization, management and query of massive multi-source remote sensing data. The method includes: subdividing the longitude-latitude space of the earth according to a specified subdivision method to obtain multiple grids; wherein, the numbers of grids corresponding to the longitude direction and the latitude direction are the same; Hierarchical; Integer encoding is performed on the coordinate information corresponding to each of the grids to obtain first coded values ​​corresponding to each of the grids; establishing the grids and the first encoding values ​​corresponding to each of the grids An association relationship between the encoded value and the remote sensing image data. The technical solution can use the global multi-scale grid integer coding to realize the location association with the multi-source remote sensing data, and realize the space-time index and query of the multi-source remote sensing data.

The present invention relates to the field of computer vision and deep learning, and in particular to a video group behavior recognition method based on cascaded Transformers. First, a video data set is collected and generated, and the video data set is extracted through a three-dimensional backbone network to extract three-dimensional spatio-temporal features, and a key frame image space is selected. Feature map; preprocess the key frame image space feature map and send it to the human target detection Transformer to output the human target frame in the key frame image; then, map the sub The feature map is combined with the frame feature map around the key frame image to calculate the query / key / value, input the group behavior recognition Transformer, and output the group-level spatio-temporal encoding feature map; finally, the group behavior is classified by the multi-layer perceptron. The invention has the effect of effectively improving the accuracy of group behavior recognition.