97 results about "Matrix sequence" patented technology

The invention relates to the technical filed of wireless communication, which discloses a transmission method, network side equipment and a terminal for signaling control, wherein, the method comprises the steps that: a linear pre-coding matrix sequence number corresponding to a resource block allocated to users by a base station is obtained; the resource block used for the transmission of a special control channel is determined, and the information of the pre-coding matrix sequence number of the resource block in which the special control channel is positioned is loaded to a physical downlink control channel; a linear pre-coding matrix which has the same data with the resource block in which the special control channel is positioned is adopted to carry out pre-coding processing to the information loaded by the special control channel. The method, the network equipment and the terminal of the invention cause PMI information corresponding to most RB to obtain performance gain by the pre-coding processing during transmission, thereby promoting the reliability of PMI information transmission.

The invention discloses a short-term traffic flow prediction method based on a generative confrontation network. The prediction method can be applied in an urban traffic road network, wherein historical traffic data is modeled into a matrix sequence, a depth model is used for learning spatial-temporal correlativity in the data, a generator and discriminator in the network are iteratively trained,the trained model can predict the traffic conditions of all road segments in the road network in a short time in the future. The short-term traffic flow prediction method based on the generative confrontation network fully utilizes the spatial-temporal characteristics of the traffic data, the predicted object is extended from a single road segment to the entire traffic network, and the accuracy ofthe prediction is significantly improved.

A construction method of a code book is applicable to a multi-input multi-output system of 2 sending antenna. The method comprises the following steps: dividing a theta k and a phi k to i, j angle values which have same rank and are quantified; respectively substituting the quantified angle values to a V for figuring out a matrix, and the V is indicated in the right; after using the matrix to divide the root of the number N of a data stream which waits for sending, and taking out front N rows of the matrix to get a codebook. After configuring the code book on a sending terminal and a receiving terminal, a pre-coding method which is based the code book manner executes the following steps: the receiving terminal conducts the analysis of singular values on a channel matrix which is acquired through the estimation of a channel so as to get a pre-coding matrix V of the sending terminal; the receiving terminal finds out a code book matrix which is close to the V matrix on the code book which is configured on the receiving terminal, and sends the sequence number of the code book matrix to the sending terminal; the sending terminal obtains a pre-coding matrix from the code book which is configured on the sending terminal according to the received code book matrix sequence number, and uses the pre-coding matrix to conduct the pre-coding on the data stream. The invention has the advantages of enabling the 2 sending antenna to get better performance under the condition of costing a little.

A method and an apparatus for predicting inter-area travel demand are disclosed. The method comprises the steps of: step S1, constructing a deep model of multi-context information extraction, utilizing a traffic demand matrix sequence of a plurality of historical time periods as an input and an actual traffic demand matrix of a corresponding time period as a target output, and utilizing a back propagation algorithm of a neural network to train the deep model; step S2, constructing a traffic travel demand matrix sequence rich in context information; step S3, taking the deep model parameters andthe deep model trained in step S1 as the final predictor together, inputting the continuous traffic trip demand matrix sequence, and predicting the unknown traffic trip demand matrix of the next timeperiod. The invention can improve the accuracy of the inter-area trip demand prediction.

The invention discloses a differential encoding space-time-frequency modulation method, and belongs to the technical field of wireless communication. The method comprises that: a mobile terminal of a communication system performs space-time encoding on an information bit sequence required to be transmitted to generate a unitary matrix code word sequence; the unitary matrix code word sequence is subjected to differential modulation to generate a matrix sequence to be transmitted; symbols in the matrix sequence to be transmitted are respectively mapped to a space domain, a time domain and a frequency domain so as to acquire a group of space-time-frequency three-dimensional signals; a plurality of groups of transmission signals are acquired through OFDM modulation, and each group of transmission signals is transmitted through a transmitter antenna; and the OFDM modulation and differential demodulation are directly carried out a receiving end, and required information is acquired. Through differential space-time-frequency block code mapping or differential space-time-frequency cyclic code mapping, the encoding and the differential modulation are respectively carried out at different dimensionalities, so that a requirement of a channel on relevance of the time domain and the frequency domain is reduced, and a requirement of reliable signal transmission in high-speed mobile environment in a wideband communication system can be met.

The invention discloses an abnormal double-person interaction behavior recognition method based on a vision co-occurrence matrix sequence, and the method comprises the steps: 1, carrying out the motion detection and segmentation of a transaction behavior in a video collected by a camera; 2, respectively carrying out the regional HOG feature extraction of left and right action performers in the video; 3, constructing a vision word through employing the HOG features extracted at step 2 and a K-means algorithm, generating a vision word bag, coding the words in the vision word bag, carrying out the vision word coding of region features through employing a similarity measuring function, carrying out the statistics of vision co-occurrence relation among the interaction individuals in a time dimension, and obtaining the vision co-occurrence matrix sequence so as to represent the abnormal double-person interaction behaviors in the video; 4, carrying out the training and recognition of an HMM algorithm. The method is simple and efficient, and is higher in recognition accuracy. Aiming at the recognition of abnormal double-person interaction behaviors in an intelligent monitoring system, the method is better in recognition performances.

The present invention discloses a device for reminding the pedestrian to cross a street safely at night. The device for reminding the pedestrian to cross the street safely at night is characterized in that a speed bump 1, a speed bump 2 and a speed bump 3 are arranged in the lanes, pedestrian induction devices are arranged in a pavement, and the three speed bumps are all induction type light reminding speed bumps; each pedestrian induction device is divided into a pedestrian waiting induction area and a pedestrian direction determination area which are both sensing coils, and the sending coils are connected with a junction box and then are connected with a control chassis. Compared with the prior art, the pedestrian induction devices determine the pedestrian crossing directions by a method of analyzing a matrix sequence, thereby improving the accuracy of the induction devices of identifying the pedestrian crossing; the deceleration reminding devices realize a simple and visual reminding effect by combining the pedestrian crossing identifiers with the vehicle speed-limiting prompt marks, by a segmented reminding mode and under the premise of not occupying the excessive road space.

The invention provides a method for detecting fMRI brain network dynamic covariations. The method comprises the following steps of: obtaining a sample; Firstly, constructing a brain functional networkby utilizing functional magnetic resonance data; according to a certain basis (factors such as age, disease course length, education degree and the like), different network matrixes are arranged in sequence;obtaining a cross-tested network matrix sequence; then, a sliding window method is used for researching a covariant relationship between ROI brain regions which cross a tested object; and obtaining a series of covariant matrixes, and depicting the difference of the brain interval collaborative activity modes under the trans-tested scale is by calculating mathematical indexes for describing covariant relationship variability or through statistical inspection, so that the interactive relationship between brain regions in the network can be better understood.

The invention relates to an interlacing method used for a WLAN frequency hopping system, and belongs to the communication signal processing technology field. An information sequence is converted into an m*n matrix, according to a matrix size of a signal sequence, a prime code matrix of an appropriate dimension is generated, according to a prime code matrix sequence the signal sequence is subjected to interlacing, and after prime code interlacing the signal sequence is converted into a signal sequence and is sent. After receiving the signal sequence, a receiving terminal converts the signal sequence into an m*n matrix with a same dimension of the sending terminal, the signal sequence is subjected to deinterlacing by using a prime code matrix which is synchronous with the sending terminal, and the signal sequence after deinterlacing is converted into a signal sequence used for post processing. The method is flexible in realization, and when a series of errors appear, error code resistance performance which exceeds error correction capability of an error correction code can be improved; the method has a good inhibition effect on single-tone interference; situations of high time-delay and large memory space in the prior art are improved, and anti-interference performance of frequency hopping communication is improved.

The invention discloses a long-term network traffic prediction method based on deep learning, and the method comprises the steps: obtaining a regional network traffic sequence, and carrying out the statistics of a traffic value used at each moment; preprocessing the traffic matrix sequence, and obtaining the input data of a Transform model; establishing a Transform model, and carrying out adaptive extraction of temporal correlation and spatial correlation on the two-dimensional matrix data by adopting the Transform model; and finally, carrying out model training by adopting a self-adaptive training mechanism. According to the method, the accuracy of multi-step long-term prediction of the network flow is improved, so that an operator can plan future network resources in advance, and the reasonable allocation of wireless resources is facilitated.

This invention relates to utilising image compression and reconstruction algorithms for transmission and reception of data that can be represented as a numerical matrix or a sequence of numerical matrices in bandwidth limited channels. The method and algorithms described herein may reduce data size to such an extent that data that was considered too large to be practically transmitted in a useful time frame can be transmitted in real or near real time. The algorithms can be applied to data communications for a variety of sensor, electronic and information systems.

The invention discloses an electroencephalogram spatial-temporal feature learning and emotion classification method based on a hybrid neural network. The method comprises the steps of collecting electroencephalogram signals of multiple channels; extracting PSD (power spectral density) features from the electroencephalogram signals of the multiple channels; converting the features into a two-dimensional mesh matrix sequence; dividing the sequence into a plurality of fragments Pj; establishing a CASC_CNN_LSTM model and a CASC_CNN_CNN model, jointly extracting deep spatial features and temporal features of the electroencephalogram signals from each fragment Pj through the CASC_CNN_LSTM model, and inputting the deep spatial features and temporal features extracted by the CASC_CNN_LSTM model into a softmax layer corresponding to the CASC_CNN_LSTM model to perform emotion category prediction; jointly extracting deeper spatial features of the electroencephalogram signals from each fragment Pjthrough the CASC_CNN_CNN model; and inputting the deeper spatial features extracted by the CASC_CNN_CNN model into a softmax layer corresponding to the CASC_CNN_CNN model to perform emotion categoryprediction. According to the invention, the emotion classification is more accurate.

The invention provides a text recognition method and device, a storage medium and electronic equipment, and the method comprises the steps: obtaining a to-be-recognized target image, and extracting afeature matrix of the target image based on a preset convolution layer; determining a plurality of sub-feature matrixes according to the feature matrix, and generating a matrix sequence containing allthe sub-feature matrixes; and taking each sub-feature matrix in the matrix sequence as input of a recurrent neural network in sequence, determining a text corresponding to each sub-feature matrix, and determining a text sequence in the target image according to all the determined texts. According to the text recognition method and device, the storage medium and the electronic equipment provided by the embodiment of the invention, the text or the character corresponding to each position is comprehensively determined based on the glyph features and the semantic features in the target image, thecharacters with similar glyph can be effectively distinguished, the text recognition accuracy is higher, and the recognition rate is greatly improved.

The invention relates to a multi-STATCOM (Static Synchronous Compensator) damping controller design method. The method comprises the following steps: (1), setting a power system which has N STATCOMs under the same oscillation mode, and sequentially setting the parameters of lead-lag links of the N STATCOM damping controllers; (2), taking each STATCOM critical gain matrix as a basic value to unify operation gain matrices so as to obtain a gain coefficient matrix; (3), calculating a gain matrix sequence of an operation gain micro-increment test; (4), obtaining a gain matrix sequence extreme value and an operation gain through a synchronous gain micro-increment test for the N STATCOM damping controllers; (5), determining the operation gain of the N STATCOMs based on gain constraint conditions; and (6), verifying the damping effect, regulating the gain, and finally determining a coordination operation gain of the N STATCOM damping controllers. Due to the adoption of the method, the potential hazards in the safety and stability of a system are eliminated so as to realize the coordinated and stable operation of the multi-STATCOM damping controllers and fully play the better damping control performance of the multiple STATCOMs, therefore, the system is improved in safety and stability.

The invention discloses a construction method of a multi-time scale time series collaborative prediction model, which comprises the following steps: 1) inputting recording data generated by the powerconsumption condition of a user with the change of time, and constructing a relevant time sequence representation according to the recording data; 2) based on the time sequence representation obtainedin the step 1), analyzing all the power consumption data, capturing different characteristics and analyzing the corresponding variation rules, and thereby constructing a time scale matrix sequence; 3) according to the output of the step 1) and the step 2), constructing a prediction model of the time series of the power consumption data, wherein the prediction model of the time series is a multi-scale RNN model; 4), according to the output of step 1), step 2) and step 3), and the output of external factors, the weighted fusion solution is carried out to obtain a multi-time scale power consumption sequential data cooperative prediction model. The method of the invention enables the accuracy of the power consumption prediction of the user to be improved.

The invention provides an action recognition method of a self-adaptive mode, and the method comprises the steps: 1, collecting a human body motion video, extracting the position information of each frame of human body skeleton point, and forming a whole body skeleton point motion sequence; 2, reading a half-body skeleton point action sequence from the whole-body skeleton point action sequence, andperforming data processing to obtain a half-body action posture matrix sequence Sp and half-body action conformity s1, and the whole-body action posture matrix sequence Sh and the whole-body action conformity s2; step 3, if s1 is not greater than s2, taking Sh as input data; and otherwise, taking the Sp as input data, and performing action recognition on the input data by adopting an SVM supportvector machine to obtain a recognition result. According to the method, different action modes are adopted for recognition, the optimal skeleton point set is selected to represent the action posture characteristics, the influence of irrelevant joint points on action quality evaluation is effectively avoided, and the accuracy of action recognition is improved.

The invention discloses a matrix multiplication operation method and device and a computer readable storage medium, and the method comprises the steps: obtaining a first scalar parameter of a to-be-operated first matrix and a second scalar parameter of a to-be-operated second matrix when a matrix multiplication operation instruction is received; determining a first matrix and a second matrix basedon the first scalar parameter and the second scalar parameter, and determining a first segmentation parameter, a first filling parameter, a second segmentation parameter and a second filling parameter; performing sub-matrix segmentation filling operation on the first matrix based on the first segmentation parameter and the first filling parameter to obtain a first sub-matrix sequence, and performing sub-matrix segmentation filling operation on the second matrix based on the second segmentation parameter and the second filling parameter to obtain a second sub-matrix sequence; and sequentiallyselecting a sub-matrix in the first sub-matrix sequence and a sub-matrix in the second sub-matrix sequence to perform multiplication to obtain a plurality of products, and accumulating the plurality of products to obtain an operation result. The matrix operation speed can be increased, and memory consumption is reduced.

The invention discloses a method and system for improving floating point matrix operation efficiency. The method comprises the following steps of: extracting sample matrixes in a floating point matrix operation to obtain a sample matrix sequence; determining an optimum calibration parameter sequence of the sample matrix sequence; and transforming calibration parameters in the optimum calibration parameter sequence and applying the transformed calibration parameter sequence into all the matrix sequence operations. By utilizing the method and system disclosed by the invention, the floating point matrix sequence operation response time can be shortened and the operation precision can be ensured.

The invention discloses a three-dimensional magnetic code based train positioning identifying method. The method comprises the following steps: combining the magnetic N pole, S pole and 0 vacancy of apermanent magnet, which represent numerical values 2, 1 and 0 respectively, to form three-dimensional magnetic codes of train positioning information; pasting the three-dimensional magnetic codes along a track according to certain interval; and performing information reading and decoding by an electromagnetic induction type identifying and reading device mounted on the train, so as to obtain coordinate positioning as well as the speed and acceleration of the train. In the method, the preset positions of the train are subjected to coded combination by using three code elements according to a matrix sequence way, and the preset sequences of different code elements represent different positional information; and the method relates to the technologies such as coding principle, inductive identification and fast decoding identification. The most remarkable characteristics of the patent are strong environmental suitability and high reliability, and the method can be used normally in varioussevere environments such as heavy dust, poor light, low temperature and rainy and snowy weather.

The invention provides a user behavior sequence anomaly detection method, a terminal and a storage medium. The method comprises the steps of obtaining user behavior information in a preset time period; aggregating the characteristic attributes according to a sequence; configuring behaviors of the user in a preset time period into row vectors, and then forming a behavior row vector time sequence; extracting behavior row vector time sequences of any two users, calculating correlation coefficients, and judging vector similarity; searching the optimal distance between the behavior row vector timesequences of the two users by adopting a dynamic warping algorithm; calculating a distance average value and a standard deviation among all users; and if the distance between the user and other usersis greater than +3 times of the standard deviation of the average value, judging that the user is an abnormal user. User behavior details can be analyzed, the problems that a feature matrix cannot begenerated and sequence lengths are inconsistent due to the fact that a user does not have continuous behaviors are solved, and the false alarm rate of anomaly detection is reduced. Therefore, abnormalbehaviors hidden in the group can be identified, and the security of data information is ensured.