1625 results about "Algorithm complexity" patented technology

The invention discloses a natural scene text identification method based on a convolution attention network; the method comprises the following steps: using a two-dimensional convolution CNN as an encoder, extracting High-level semantic features of an input image, and outputting a corresponding feature graph to a decoder; using a one-dimensional convolution CNN as the decoder, combining an attention mechanism to integrate the High-level semantic feature formed by the encoder with a character level language model, thus forming a decode character sequence corresponding to the input image. For asequence with the length of n, the method uses the CNN modeling character sequence with a convolution kernel as s, and only O (n / s) times of operation is needed to obtain the long term dependence expression, thus greatly reducing the algorithm complexity; in addition, because of convolution operation characteristics, the CNN can be better paralleled when compared with a RNN, thus performing advantages of GPU resources; more importantly, a deep model can be obtained via a convolution layer stacking mode, thus improving abstractness expression of a higher level, and improving the model accuracy.

The invention relates to a position / force hybrid control method for a space manipulator, which is a double-loop control method combining a Cartesian space and a joint space. Compared with a conventional control method on the basis of an integral model, which has the problems of oscillation of the transition process and algorithm complexity, the position / force hybrid control method for the space manipulator has the advantages that the position / force hybrid control problem is decomposed into the planning problem of the Cartesian space and the control problem of the joint space; a position / force kinematic plan from the Cartesian space to the joint space and kinematic control of the joint space are given out; an internal and external loop control system is formed; and position control on a free space at the tail end of the manipulator and contact force control in a space restrained by the environment are implemented. A simulation result for a SDOF ( Seven Degrees of Freedom) redundant manipulator proves that the method disclosed by the invention is simple to operate, is effective and is easy for engineering realization.

The invention discloses a mode selection and resource allocation method for device-to-device (D2D) users in a cellular system, which reduces the algorithm complexity of the mode selection and resource allocation for the D2D users, and the obtained mode selection and resource allocation for the D2D users can reduce the transmission power needed by the D2D users. The method comprises two steps of mode and resource demand calculation and analysis and relevant algorithm for the resource demand and supply relation. The mode and resource demand calculation step comprises the following steps of calculating the minimal transmission power needed by each pair of D2D users under a cellular mode, a special channel mode and a multiplying channel mode; calculating the power difference of the D2D users under the three different modes; calculating the mode demand and the resource demand of each pair of D2D users; and counting the sum of the resource demand of the system. The analysis and relevant algorithm step for the resource demand and supply relation comprises the steps of further obtaining a mode selection and resource allocation algorithm under each situation according to the resource demand and supply relation of the system.

The invention discloses a multi-band structure self-adaptive filter switching method for AEC (acoustic echo cancellation). Firstly, a far-end voice signal is acquired; a voice endpoint is detected, and a VAD (voice activity detection) flag bit and an improved envelope decision threshold are output; the voice signal is fed into a loudspeaker to serve as a desired signal and also input into a self-adaptive filter; the self-adaptive filter adopts a switchable multi-band structure and a corresponding self-adaptive algorithm, parameters of the filter are adjusted by use of the least mean square criterion according to feedback information, and the optimal solution is obtained. According to the provided switching method, voice characteristics are considered sufficiently under the condition that steady maladjustment is guaranteed, and optimized configuration of the convergence rate and the algorithm complexity is realized while advantages of the algorithm in the convergence rate are utilized. During actual application of echo cancellation, a single algorithm does not easily meet various variable demands. The variable switching algorithm provides more probability for a user and has great significance in application of self-adaptive echo cancellation.

The disclosed multi-classifier combination face recognition method based on feature sorting comprises: extracting face area from initial image for pre-process and feature extraction; feature sorting to obtain different face feature groups; designing component classifier for every group to recognize face and combine results for optimal effect. This invention overcomes dimension disaster, reduces algorithm complexity, and improves recognition performance.

The invention discloses a SVM (Support Vector Machine)-based multi-sensor target tracking data fusion algorithm. According to target information acquired by sensors, a compact combination mode is adopted, the SVM serves as a middle layer, an environmental variable and a measurement variance normalized vector serve as input of the SVM, output of the SVM serves as trust of the sensor, a known training sample is used for offline training, real-time filter information is used for online estimation, and a fusion knowledge base performs track fusion through real-time weighting according to the obtained trust of the sensor. According to the SVM-based multi-sensor target tracking data fusion algorithm and the system thereof, the SVM principle is adopted, the algorithm complexity is low, the environmental variable and the measurement variance normalized vector are introduced, the biological robustness and the fault tolerance are strong, extension is easy, and the algorithm and the system thereof are applicable to the field of multi-sensor tracking.

The invention provides a micro-gesture identification method and system based on ultrasonic active detection. The method includes the following steps of emitting an ultrasonic signal and receiving an ultrasonic echo signal, reflected by an identified gesture action, of the ultrasonic signal; comparing the ultrasonic signal with the ultrasonic echo signal to obtain the reception time delay and frequency resolution variance of the ultrasonic signal and the ultrasonic echo signal; according to the reception time delay and frequency resolution variance, obtaining the distance and velocity information of the timing sequence in the ultrasonic echo signal; and according to the distance and velocity information, performing identification and determination on the identified gesture action through a data driving pattern recognition algorithm. The micro-gesture identification method and system can accurately identify the micro-gesture actions and have the advantages of high resolution, high robustness, low consumption, and low algorithm complexity.

The invention discloses a posture identification method based on near-infrared TOF camera depth information. The posture identification method includes the steps that a depth image is obtained through an SR4000, earlier-stage operation processing is carried out on the depth image, a human body object is obtained and skeletonized, then coordinates of end points of the head, the hands, the feet and the mass center of the human body target are obtained, the heights between the end points and the ground and the skeleton line distances between the end points and the mass center are calculated, the attributes of the end points are identified according to the skeleton line distances and the ground heights, and then posture identification is achieved in cooperation with motion characteristics and the depth information. In this way, the target tracking difficulty and the algorithm complexity are lowered, the identification effective rate and the target division accuracy are improved, and convenience is brought to real-time embedded transplanting. The invention further discloses a device used for implementing the posture identification method.

The invention discloses a method for synchronizing audio / video data in a real-time video monitoring system. The method comprises the following steps of: acquiring acquisition time of audio data and video data by a monitoring end, encoding the audio data and the video data, and packaging the acquisition time of the audio data and the video data in respective data packets to be sent out; and acquiring the data packets and analyzing and decompressing the data packets by a client, and decoding and playing the data packet of the audio data and the data packet of the video data according to a synchronization mechanism. According to the invention, after the client acquires data transmitted by the monitoring end, the data packet of the video data and the data packet of the audio data are respectively put in a video cache queue and an audio cache queue according to a time sequence, and the data are decompressed and played according to a synchronization precision requirement. The method disclosed by the invention has the advantages of greatly reducing the algorithm complexity of audio / video synchronization without reducing performances to much and completely meeting requirements for video monitoring.

The invention discloses a method and system for locating a mobile terminal, and belongs to the field of wireless service supporting in mobile communication. The method integrates the advantages of a support vector machine and a K-nearest neighbor algorithm, and through the stages of off-line drive test sampling and model training, two layers of SVM locating models and KNN locating models are constructed. In an on-line locating stage, the two layer of the SVM locating models and the KNN locating models are combined, so that the locating precision of the mobile terminal is greatly improved. The method achieves compromise and balance between the algorithm complexity and the locating precision, and compared with a traditional locating method, the method effectively improves the locating precision, greatly reduces calculated amount and improves the locating speed.

The invention provides a method and device for detecting lost articles in security monitoring, belonging to the field of video monitoring. The method comprises the following steps of: dividing a scene plane into a plurality of regions; carrying out two-frame background storage on a certain relevant region of the plurality of regions and performing subtraction within a certain interval; judging whether a target lost article exists in the relevant region according to the subtraction result, and judging whether a found article is the target lost article; if the target lost article exists, extracting the lost article and alarming; and if no lost article exists, continuously monitoring other regions. By adopting a moving detection and two-frame background subtraction method, the invention avoids algorithm complexity increase caused by incapability of accurately tracking or stricter requirement for a tracking strategy when a moving object moves too fast or background updating speed is overhigh and even the object is distorted in order to meet a hardware requirement is avoided, and erroneous judgment caused by temporarily shielding the object is also avoided.

The invention belongs to the field of panoramic stitching and provides a 360-degree stitched-type panoramic camera with a wide view field in a vertical direction. The 360-degree stitched-type panoramic camera comprises a plurality of identical image sensors which are vertically arranged, wherein the models, the number and the relative positions of the sensors are determined according to the demands for the size of a view field in panoramic stitching application. The invention additionally provides a calibrating method and a stitching method based on the panoramic camera. Through calibration, images in all loops are projected on the same plane; then the stitching of panoramic images is completed by adopting a dynamic planning method and a gradient domain fusion method. Since a wider vertical view angle can be obtained under the situation of the arrangement of a single row of sensors and an image stitching algorithm is only required to be operated horizontally, the design of the stitching algorithm can be simplified and the algorithm complexity is reduced. Since the stitching operations of adjacent images in a group are mutually independent, multiple processors can be adopted for execution, the real-time stitching of loops of videos can be realized on an FPGA (Field Programmable Gate Array) and the market prospect is wide.

The invention belongs to the field of signal processing, aims to solve the problem that a traditional DOA (direction of arrival) estimation algorithm is large in sampling data volume to cause high computation complexity, and provides a beam-domain-based multi-measurement-vector underdetermined system regularization focusing solving (BS-RMFOCUSS) algorithm on the basis of a compressed sensing theory by using airspace sparsity of target signals. According to the algorithm, a target compressing signal is mapped to a beam domain from an element domain, the shortcoming that a sparse reconstructing algorithm cannot be used under the condition of low signal to noise ratio is overcome to a certain degree, and the algorithm complexity is low. Numerical simulation shows that the property of the algorithm is superior to that of the traditional DOA estimation algorithm, and the algorithm has high angular resolution and estimation precision, and can carrying out effective DOA estimation on coherent signals.

The invention discloses a system and a method for controlling the power consumption of a mobile terminal. The method comprises the following steps of: detecting working temperature and / or the working current of a power interface, and judging whether the working temperature exceeds the maximum temperature threshold or not and / or whether the working current exceeds the maximum current threshold or not; and if detecting that the working temperature exceeds the maximum temperature threshold and / or the working current exceeds the maximum current threshold, reducing the power consumption of a radio-frequency power amplifier or reducing the power consumption of a base-band part. The system and the method solve the problem that the power consumption of a broad-band mobile communication terminal cannot be effectively controlled on various application occasions, and adjust the power consumption of the terminal in real time by detecting the working current and the working temperature without adding the hardware structure of the terminal or increasing the software algorithm complexity so as to meet the requirements of the limit on the maximum current of various interface protocols and the limit on the shell temperature of various terminal structures.

The invention discloses a camera calibration method and device for a visual navigation unmanned aerial vehicle (UAV). The method comprises steps of: grouping two adjacent monocular cameras in a multi-camera visual navigation UAV as a binocular camera group, and determining the relative position relation of respective monocular cameras in the binocular camera group; constructing a global SLAM map of the binocular camera group by using a binocular camera mechanism and determining the conversion relation of each binocular camera group relative to a reference coordinate system; merging maps obtained by the grouped cameras into a complete map; and optimizing the map to obtain an optimized output parameter. The method and the device are simple and easy to implement, and avoid a complex calibration block selection process. A continuous global map established in the calibration process can be used for the visual positioning and navigation of the UAV, reduces algorithm complexity of system real-time computing, and simplifies the operation procedure of a traditional calibration method.

The invention belongs to the technical field of communication, and relates to an interference type identification method. The interference type identification method provided by the invention can comprise the following steps: carrying out interference detection and interference identification on hostile interference in a wireless communication system in a complex electromagnetic environment; and then jointly deciding the condition of the current electromagnetic environment, and classifying the interference types in an interference environment, the existence of interference is mainly determined from two aspects of time domain and frequency domain in the interference detection; in the interference identification, feature parameters are mainly extracted from the time domain, the frequency domain and the time-frequency domain, the interference is divided into time domain interference, frequency domain interference and sweep frequency interference by using a decision tree, and a part of interference is identified in detail; and the feature parameters extracted by the interference type identification method provided by the invention have relatively good robustness and are not sensitive to the interference and noise ratio, and relatively good interference identification performance can be obtained when a decision tree judgment method having relatively low algorithm complexity is used.

The invention discloses a video keyframe extraction method, which comprises the steps of: using a ViBe algorithm fused with an inter-frame difference method to perform moving object detection on an acquired original video sequence, so as to obtain a key video sequence containing a moving object; performing keyframe crude extraction on the key video sequence by using a global characteristic peak signal-to-noise ratio to obtain candidate keyframe sequences; and establishing global similarity of the candidate keyframe sequences by using the peak signal-to-noise ratio, establishing local similarity of the candidate keyframe sequences by using SURF feature points, and performing weighted fusion on the global similarity and the local similarity to obtain comprehensive similarity, performing self-adoptive keyframe extraction on the candidate keyframe sequences by using the comprehensive similarity, and finally acquiring a target keyframe sequence. The video keyframe extraction method providedby the invention can effectively extract the video keyframes, obviously reduce the redundant information of video data, and express the main content of the video concisely. Moreover, the video keyframe extraction method has low algorithm complexity and is suitable for real-time extraction of keyframes of surveillance videos.

The invention relates to a speech coding method based on multiple description of grid vector quantization, which belongs to the field of communication technology and is mainly applied for group speech transmission. The method first of all conducts LPC analysis on speech signal, carries out perception weighting and transfers the weighted speech signal to a frequency domain. The signal is decomposed into vector in the frequency domain and is quantized with grid vector quantization method. Finally, the quantized grid point is decomposed into two descriptions (index of corresponding sub-gird point) according to the marking method of multiple description grid vector quantization and is transmitted in a signal channel, so as to enhance the transmission reliability of the system. Because this method is based on grid quantization technology, the complexity and encoding time delay are much lower.

The invention relates to a method and a device for classifying network traffic on line based on decision tree high-speed parallel processing. The method comprises the following steps of: performing acquisition, distribution and manual classification on early real traffic data, extracting the packet characteristics of an early transmission control protocol (TCP) stream set, establishing a decision tree classification model, converting a data structure, performing distribution and class judgment on a data packet to be classified, tagging a current data packet, extracting the packet characteristics of a TCP stream to be classified, and searching for a decision tree. The device comprises a decision tree construction module, a structure conversion module, a classification result processing module, a medium access control (MAC) layer processing module, a data packet polling management module, a distribution judgment module, a traffic information extraction and tagging module, and a decision tree searching module. The method and the device are low in algorithm complexity and high in processing speed, classification accuracy and stability, and can be used for equipment and systems with requirements for online traffic classification in a high speed backbone network; and online classification can be realized.

The invention relates to an RSS (received signal strength) fingerprint database based secondary fuzzy clustering indoor-positioning method which includes the following steps of firstly, setting a position of an indoor signal coordinate node and forming an indoor positioning system; secondly, selecting a position of a reference point, measuring an RSS vector accepted by the reference point, and setting up an RSS fingerprint database; thirdly, clustering primarily to obtain a point same with a point to be positioned; fourthly, obtaining a reference point half similar to the point to be positioned, and obtaining an overall nearest point; fifthly, clustering secondarily to obtain a point same with the overall nearest point; and sixthly, weighting all points according to similarity, and computing coordinates of the point to be positioned. The RSS fingerprint database based secondary fuzzy clustering indoor-positioning method has the advantages that impact on positioning accuracy from multiple paths and non-line-of-sight and the like can be avoided effectively, no extra facility support is required and positioning algorithm complexity and cost of a positioning system are reduced.

The invention discloses a method for implementing a stunt mode function of a transport stream file based on an indexed file. The method is characterized by comprising the following steps of: acquiring related information of a transport stream (TS) file, and writing the information in the head of the indexed file; creating an index descriptor; searching I frame data; acquiring position offset of the I frame data in a source file and a display time stamp of the frame, and generating the indexed file by combining the information; performing fast forward / fast reverse or positioning operation through the TS file; and closing the indexed file, and releasing a resource applied for. The method for implementing the stunt mode function of the transport stream file based on the indexed file has the advantages of lower algorithm complexity, shorter time consumption, capability of accurately positioning at a corresponding position in the TS file and the like. The method is suitable to be widely popularized in the field of videocast.

Disclosed is a video quality evaluation method, which comprises: obtaining parameter information about video data, the parameter information comprising a code rate, a frame rate and data packet information; according to the code rate or the frame rate, calculating the video reference quality of the video data; according to the data packet information, calculating the number of continuously lost data packets, and according to the number of continuously lost data packets and a first data packet threshold, calculating the number of times of effective packet loss; according to the number of times of effective packet loss, calculating the video packet loss distortion quality of the video data; and according to the video reference quality and the video packet loss distortion quality, calculating the final video quality. The number of times of effective packet loss can be applied to accurately evaluate a video more effectively, and the algorithm complexity is low. Also disclosed is a video quality evaluation device.

The invention discloses a low-complexity hybrid precoding method in a millimeter wave large-scale MIMO system. In the method, the minimum F norm is got between corresponding positions of a precoding scheme corresponding to each radio frequency link and partially connected with a hybrid precoding scheme and an optimal digital precoding scheme under the condition of not considering the radio frequency link, and thus a digital precoding matrix and an analog precoding matrix of each radio frequency link are got via computation. The phase of the analog precoding matrix of each radio frequency linkis the phase of a maximum singular value vector of the optimal digital precoding matrix, the real part of a digital precoding matrix element is the product of conjugate transpose of the analog precoding matrix and the real part of an optimal digital precoding vector at a corresponding position, and the imaginary part of the digital precoding matrix element is the product of the conjugate transposeof the analog precoding matrix and the opposite number of the imaginary part of the optimal digital precoding vector at the corresponding position. Compared with analog precoding, the method providedby the invention can acquire the higher system accessibility and speed with the lower algorithm complexity.

The invention discloses a vehicle detecting method based on Gauss difference multi-scale edge fusion. The method includes the steps that Gauss scale transformation is performed on images to obtain four Gauss images in adjacent scales; according to the four Gauss images in the adjacent scales, the difference operation is performed between the images in the adjacent scales to obtain three Gauss difference images different in scale, edge detection is performed on the obtained three Gauss difference images through a Sobel operator, then edge fusion with the scale upward searching is performed to remove a lot of background edges while edge information of a vehicle is obtained as much as possible, and expansion, closed operation, hole filling and other series of morphological operation are performed on the fused edge images to obtain a connected domain image representing the vehicle; an outside rectangle of the position where the vehicle is located is determined in the original image according to the position information of a connected domain to detect the vehicle. The images in multiple scales are processed, so that algorithm complexity is reduced, operation amount is reduced, efficiency of vehicle detection is effectively improved, and a good detection result is obtained.

The invention publishes a medical image ROI compression method based on a lifting wavelet and a PCNN, which comprises the following steps of: circling a region of interest by a doctor, and separating the region of interest from a region of no interest by a difference image method; adopting lossless compression in the region of interest, constructing compactly supported biorthogonal wavelet transformation through a lifting scheme, and then carrying out Huffman encoding; adopting lossy compression in the region of no interest, segmenting gray-value pixel approximate points through the PCNN, carrying out ignition operation, and then carrying out run-length encoding; and finally carrying out inverse transformation restoration, merging the region of interest and the region of no interest, and eliminating a boundary discontinuity problem through linear interpolation. Experimental results prove that the region of interest can be flexibly selected and controlled by the compression method, the used information for doctor diagnosis can be completely reserved, and the compression ratio is higher. Meanwhile, the computation for an ROI mask and the computation and the encoding for a wavelet coefficient difference value are omitted, the compression and decompression time and the algorithm complexity are reduced, and the image processing and transmitting efficiency is improved.

The invention relates to a heterogeneous wireless network access selection method based on users. According to different service flow communication requirements of the users, network, terminal and user parameters are extracted, parameter key factors are quantized based on a fuzzy function, parameters are compared through an analytic hierarchy process, and the weights of the key factors are obtained; then according to an ordinal number preference algorithm similar to an ideal scheme, the close degree value of alternative network parameters is obtained by combining the parameter quantized value with the weights; network access is selected according to the close degree value. Network received signal intensity is sampled and pre-estimated, and calling drop caused by attenuation of received signal intensity of a selected access network is reduced. Through an improved fuzzy decision quantitative method, algorithm complexity caused by building of a fuzzy rule base is avoided, and calculation complexity is lowered. Optimal network access can be selected according to service types of the users, network congestion is reduced, and network resources are fully utilized.

The invention relates to the technical field of computer networks, and provides a virtual network resource allocation method. The method comprises the steps of: building a resource allocation model; submitting competitive demands to an InP (Internal Network Processor) through SPs (Service Providers); calculating the quantity and consumption of the resource through the InP; acquiring the resource, assuming the consumption and calculating the income through each SP; adjusting the competitive strategies through each SP; and waiting for competing at the next time based on the adjusted competitive strategies. In the scheme of the invention, a novel virtual network resource allocation method is provided; the resource is periodically allocated to a plurality of service providers based on the trend of the demands on the virtual network resource, so that the resource allocation can be carried out as demands, and the performance of the whole network can be improved. Moreover, a scheme for effectively selecting the competitive strategies is provided at the same time; by using the scheme, the service providers can be guided to select the competitive strategies; the real network demands can be quickly obtained; the network performance can be improved; and the complexity of the algorithm can be effectively reduced.

The invention discloses an SAR (Synthetic Aperture Radar) image change detection method based on neighborhood similarity and mask enhancement, relating to the field of remote sensing image processing. The method mainly solves the problem that a difference map construction of SAR image change detection is seriously influenced by SAR image speckle noise and comprises the following steps of: (1) geometrically correcting and registering two SAR images I1 and I2 in the same scene at different times; (2) constructing a difference image map DI of the images I1 and I2 by using a neighborhood similarity operator; (3) performing mask enhancement on the difference image map DI to obtain a new difference image map NDI; and (4) clustering the new difference image map NDI to obtain a change detection result map CDI by using a K mean value clustering method determined by an initial point. The method has the characteristics of high operation efficiency, low algorithm complexity, low time complexity and good experimental effect, and can be used for the change detection of two time-phase SAR images.