110results about How to "Avoid computational complexity" patented technology

The invention discloses a voice quality enhancement system. The voice quality enhancement system comprises a voice decoder, a time-domain self-adaptation gain control module, a target gain real-time calculation module, a voiceless sound frequency spectrum adjustment module, a voiced sound frequency spectrum adjustment module, an environmental noise collection and estimation module and an audio output module, wherein the voice decoder, the target gain real-time calculation module, the voiceless sound frequency spectrum adjustment module and the voiced sound frequency spectrum adjustment module are connected with the time-domain self-adaptation gain control module, both the target gain real-time calculation module and the voiced sound frequency spectrum adjustment module are connected with the environmental noise collection and estimation module, and both the voiceless sound frequency spectrum adjustment module and the voiced sound frequency spectrum adjustment module are connected with the audio output module. According to the voice quality enhancement system, comfort and intelligibility of voices output by a voice output device in different environment are improved.

The invention discloses a static contact angle calculation method in the technical field of material performance test. The method includes: simulating to generate droplet margins of different sizes and contact angles on the basis of the Young-Laplace equation; calculating the contact angles of the droplet margins generated by simulation to obtain relation among the calculated contact angles, the droplet sizes and the real contact angles; taking a picture of a real droplet image and calculating the contact angle of the image; and using the interpolation or similar methods to obtain accurate contact angles according to the contact angle of the actual image and the relation among the calculated contact angles, the droplet sizes and the real contact angles. Errors brought by large droplet sizes and contact angles to the circle fitting method and theta/2 method can be effectively reduced, calculation efficiency is improved, and programming difficulty is lowered.

The invention belongs to the field of concrete compressive strength prediction, and particularly discloses a method for predicting concrete compressive strength based on a random forest and an intelligent algorithm. The method comprises the following steps: establishing an original sample set of a concrete compressive strength index system, taking a training number set as the input of a random forest regression model to carry out importance evaluation on influence factors forming the concrete compressive strength index system, carrying out feature selection, and selecting an influence factor set with the minimum error as an optimal feature variable set; training the least squares support vector machine model by taking the test data set as the input of the least squares support vector machine model and taking the concrete 28d compressive strength value as the output, and verifying the prediction result of the trained least squares support vector machine model by adopting the test data set; and performing error analysis on a prediction result. The invention improves the precision of the prediction model, enables the prediction result to be more accurate and stable, and can serve as an effective tool for quickly predicting the concrete compressive strength.

The invention relates to a high-precision rapid filtering and smoothing algorithm of GNSS/SINS tight combination. An approximation method including one-step prediction of a state equation is used to optimize the calculating efficiency of Kalman filtering; a GNSS state parameter is rejected and an SINS state parameter is decomposed to improve the calculating efficiency of an RTSS and FBC smoothing algorithm; RTSS is embedded into FBC, a new smoothing algorithm is obtained via fusion, and the resolving precision of the smoothing algorithm is improved; and a filtering result is stored in a low-frequency observation update rate, smoothing is carried out, memory consumption is reduced, and an SINS mechanical layout and high-frequency interpolation method is used to recover a tight combination solution of a random time from a low-frequency smoothing result. Thus, the GNSS/SINS tight combination can generate a high-precision resolution result rapidly.

The invention discloses a mobile context point cloud simplification algorithm based on a point feature histogram (namely a PFH operator). The algorithm mainly comprises the steps of receiving a point cloud set from a point cloud obtaining device; calculating the PFH operator of each point; calculating the standard deviation of the PFH of each point, and designing simplification rules; calculating out a corresponding relation table between the threshold values and the simplification rates of different standard deviations; according to the requirements to the simplification rate from a user, determining the threshold value; reserving the point sets with standard deviations less than or equal to the threshold value in point cloud data, in point cloud data, and eliminating other point sets. According to the mobile context point cloud simplification algorithm based on the point feature histogram, the simplification rate is high, the realtime performance is great, and adequate geometrical characteristics are reserved for scene reconstruction.

The invention discloses a space non-cooperative target pose measurement method based on a model. The space non-cooperative target pose measurement method is characterized by comprising the following main steps: obtaining a deep edge and a texture edge of a target model according to a relative pose at the last moment so as to obtain a visible edge model; extracting a Harris characteristic of a current scene image according to a KLT (Karhunen-Loeve Transform) algorithm; projecting the target visible edge model in an image scene, searching an image edge in a normal direction along the edge model and calculating a residual error of a corresponding characteristic point pair; estimating 3D coordinates of the Harris characteristic of a last frame through deep buffer, projecting the 3D coordinates in a current image in a current estimation pose, and calculating a residual error of a corresponding characteristic point pair; determining weights of characteristic point pairs according to an M estimator robust method and distributing the weights of characteristic points according to the average residual error obtained in the two steps; and iteratively calculating a relative pose according to a visual servo method. According to the method provided by the invention, the measurement of a relative pose of a space non-cooperative target can be realized, and the advantage of low calculating amount is achieved.

The invention discloses a UAV heterogeneous network multi-dimensional resource dynamic management method, and aims to provide a network management strategy method capable of reducing the calculation complexity and improving the generalization of a learning model. According to the technical scheme, a unified federated learning architecture is composed of a ground base station centralized convergence layer and a UAV distributed execution layer, and a ground base station computing platform constructs a multi-agent enhanced learning model based on any UAV heterogeneous network scene and initializes model parameters; the distributed execution layer outputs a multi-dimensional resource management behavior by using a multi-agent reinforcement learning algorithm, obtains rewards and state transition feedback of a network environment to agent behaviors, and uploads model parameters to a local base station associated with the distributed execution layer. And the ground base station obtains all local model parameters through interaction and issues the updated parameters to each UAV learning model. According to algorithm stop conditions, a unified federated learning framework outputs model parameters capable of compromising and optimizing local and global performances.

The invention relates to a three-dimensional target detection system based on laser point cloud; the system comprises a voxel size division module, a feature coding module, a feature extraction and fusion module, a target regression and detection module and a laser radar. The output end of the laser radar is connected with the input end of the target regression and detection module through the voxel size division module, the feature coding module and the feature extraction and fusion module in sequence, and during use, firstly, the voxel size division module performs voxel division on a three-dimensional target point cloud obtained from the laser radar by adopting different voxel scales; a plurality of voxelized point clouds are obtained, then feature coding is performed on the plurality of voxelized point clouds by a feature coding module; feature extraction and fusion are performed on the coded voxelized point clouds by a feature extraction and fusion module to obtain a final feature map; and finally, a three-dimensional target detection box is obtained by a target regression and detection module according to the final feature map. The design can guarantee that the structural features of the point cloud are not lost, and improves the detection precision of the three-dimensional target.

The invention discloses a method for recognizing falling of old people and an accompanying robot, and relates to the field of falling posture recognition. The method comprises the steps: 1, collectingbody video data; 2, positioning a human body; 3, acquiring skeleton information based on the result; 4, calculating a head and foot height difference value hd, a head and foot height difference valuehp of a front frame and a rear frame, and a speed v and a shoulder width w corresponding to the head height difference value of the front frame and the rear frame according to the skeleton information; and 5, establishing a judgment formula according to the data, outputting a result, calculating a distance measurement error, judging whether the distance measurement error meets a dichotomy condition or not, if so, identifying by adopting the established neural network model, and if not, returning to the step 1. According to the method, figure extraction is carried out firstly, and then posturerecognition is carried out, and dynamic and static recognition is carried out through skeleton information, so that the problem of low recognition accuracy caused by local feature extraction and detection in the prior art is solved, and the falling behavior of the old people can be correctly and rapidly recognized.

The invention provides a ball head positioning device for positioning irregular workpieces in a combined clamp and a positioning method of the ball head positioning device. The ball head positioning device comprises a ball, a boss and a cylinder which are connected in sequence and of which center lines are superposed. The positioning method includes steps of firstly simulating positioning by a computer, namely simulating to arrange a workpiece at a correct position and arrange the ball head positioning device near a form surface to be positioned; keeping the ball surface of the ball head positioning device in point contact with the form surface to be positioned of the workpiece by contacting function of a computer three-dimensional design system; measuring coordinate sizes of various surfaces of a clamp base body away from the ball center point of the ball head positioning device by means of a computer three-way measuring function after simulating positioning of the computer is finished, then adjusting workpieces positions of a combined clamp fittingly connected with the ball head positioning device, and finally accurately positioning the workpiece. The ball head positioning device solves the problem that conventional combined clamps cannot position irregular workpieces.

The invention belongs to the technical field of concrete structure impermeability prediction, and particularly discloses a method for predicting durability of a concrete structure based on a random forest and an intelligent algorithm. The method comprises the following steps: constructing a concrete impermeability index system, establishing an original sample set, taking the training number set asthe input of a random forest regression model; carrying out importance evaluation on influence factors of the impermeability index system; selecting an influence factor set with the minimum error ofthe random forest regression model; taking the optimal characteristic variable set as an input variable of the least squares support vector machine model, taking a concrete chloride ion diffusion coefficient prediction result as an output variable, training the least squares support vector machine model, and then verifying the prediction result of the trained least squares support vector machine model by adopting the test number set; and analyzing the prediction result, and verifying the effect of predicting the impermeability of the concrete structure by the least square support vector machine model. The method is high in prediction speed and accurate and reliable in prediction result.

The invention discloses a voice tampering positioning detection method based on a quantitative characteristic. The intensive study can be carried out based on the MP3 tampering position detection algorithm of the frame deviation of the prior art, and the updated detection method is provided for the description of the quantitative characteristic. By adopting the detection method, the frame deviation detection is not required, and then the complicated complexity caused by the frame deviation calculation can be prevented, and then the calculation complexity can be reduced greatly, and at the same time, the high bit rate condition of the inability of the frame deviation detection can be solved, and according to the experiment result, the detection method is provided with the higher detection accuracy.

The invention discloses a sequencing statistical decoding method assisted by partial cyclic redundancy check (CRC). The method is suitable for soft decision decoding of low-density parity check (LDPC)codes with information bits meeting CRC. Before decoding, a system generation matrix is formed according to a generation polynomial of cyclic redundancy check, the matrix is divided into two parts, one part is used for detecting a decoding result, and the other part is used for assisting decoding. A part of the CRC generation matrix of auxiliary decoding is multiplied by the generation matrix ofthe low-density parity-check code to obtain a joint generation matrix. In the decoding process, an iterative sorting statistical decoding (BP-OSD) method is mainly adopted, and a matrix is generated jointly to serve as a coding matrix in an OSD algorithm. After decoding is completed, CRC check is conducted on the information bits of the code words, and if the CRC check is met, the code serves as decoding output; otherwise the code is decided as decoding failure. According to the method, the error correction performance of the LDPC code is greatly enhanced while the CRC error detection functionis reserved.

The invention relates to a Gaussian scale space-based method for constructing semantics of a remote-sensing image. The method comprises the following steps: firstly constructing Gaussian scale space of the remote-sensing image; then respectively constructing two-dimensional Gaussian scale tree-form hierarchical structures in different space regions of the remote-sensing image under the same scale and different scales in the same space region of the same remote-sensing image; and respectively constructing nonstructural semanteme and structural semanteme based on the two different Gaussian scale tree-form hierarchical structures, and finishing construction of the semantics of remote-sensing image based on the Gaussian scale space.

The invention discloses a neural-network-based adaptive finite time command filtering backstepping control method. The method comprises the steps: constructing a finite time command filter; constructing an adaptive updating law based on neural network approximation; constructing a finite time error compensation mechanism, and constructing a dynamic auxiliary system for input saturation and the like. With the disclosed method, a problem of calculation complexity of the traditional backstepping method is solved and the errors generated in the filtering process are further eliminated. In addition, in order to further improve the robustness of the system, the method adopts a neural network approximation technology to approximate an uncertain dynamic model in the system. Besides, a dynamic auxiliary system is designed to compensate input saturation by considering the input saturation problem encountered by the actuator in the actual application process, so that the method is more suitable for actual application and a joint position tracking error is converged to an origin neighborhood which is small enough within limited time.

The invention relates to a computer vision technology, and discloses a single-image crowd counting method based on depth information and scale perception information, which improves the prediction capability and reduces the calculation complexity. The method comprises the steps of S1, performing Gaussian mapping on human head center coordinate data corresponding to an input sample picture to generate a preliminary true value density map, and correcting the preliminary true value density map based on depth information obtained by a depth estimation algorithm to obtain a true value density map;s2, predicting a crowd density map of the input sample picture by adopting a density estimation network to generate a predicted density map, calculating a loss error according to the predicted densitymap and a true value density map, adjusting network parameters through gradient back propagation, and generating a density prediction model through iteration; and S3, when crowd counting is carried out on a single image, a density prediction model is utilized to generate a prediction density map of the image, and the total number of people in the image is obtained through calculation.

The invention relates to a high efficiency data aggregation method in a smart power grid based on multidimensional data, comprising that: n users with I-dimensional data are set; an aggregator selects the generator g of big prime numbers p and Zp and overspeed incremental vector -> alpha and sends to all users; the aggregator randomly divides data p (p-1) into n uneven parts and sends the n parts ri to the user ui; after the user receives the ri, the user determines Ci according to its private data; after the user ui calculates the Ci, the Ci is transmitted to a second user by a first user; after the second user receives the data, adding and model combination are carried out to the data and the C2 of the second user until the last user sends the calculation result S to the aggregator; after the aggregator receives the result S, a recovery aggregation algorithm is executed so as to recover the aggregation data value of every dimension; the aggregator calculates according to the recovery aggregation algorithm to obtain data (D1, D2, and the like, DI) and sends the data to every user. According to the invention, the aggregation problem of massive private data is solved; the aggregation complexity of the data is reduced; the time is saved; the operability is strong; and the method is easy in realization.

The invention belongs to the technical field of capacity planning for electric taxi charging stations, and relates to a capacity planning method of the electric taxi charging stations through charging demand forecasting by modeling an operation process, in which the destinations of the electric taxis are selected based on hierarchical probability. An efficient operation region of the electric taxis is divided into multiple basic cell units with the same sizes; the similar cell units are combined into a large region; the cell units and the corresponding large region are endowed with attraction degree of pollution mobility so as to determine the destinations of the electric taxis based on hierarchical probability; and the requirement can be satisfied, and the planning capacity of the charging stations is determined according to the charging demand on each charging station by the electric taxis in the operation process based on probability. By adoption of the method, the selection efficiency for the destinations of the electric taxis can be improved; the way of hierarchical selection based on the regional characteristics can be in accord with the decision-making habits of users in a better manner, so that the problems of over-fine fine granularity, high operational complexity, and low effectiveness of the calculation effect are avoided; and convenience and rapidness in the actual operation are improved.

The invention relates to a peak-to-average power ratio inhibition method of phase shift for OFDM grouped subcarriers in a time domain. The method includes the steps that frequency domain parallel baseband signals after mapped are grouped, then IFFT calculation is conducted separately to acquire grouped subcarrier time domain equivalent basebands, and summation is conducted to acquire 4N time domain sampling values of OFDM signals; if the peak value is less than a preset limiting threshold, direct output is conducted; if the peak value is greater than the maximum of the limiting threshold, initialization of v = 2 is conducted, and the next step is implemented; a cyclical phase shift of p sampling points is conducted on the 4N sampling points of a vth grouped subcarrier according to a cyclical function, and then the next step is implemented; the peak value of the OFDM signals is detected, if the peak value is less than the limiting threshold, direct output is conducted, and if the peak value is greater than the limiting threshold, the last step is implemented in a splitting mode; when v = V, v = 2 and p = 2*p; when v <= (V-1), v = v+1 and p remains unchanged; the last two steps are cyclically implemented, and when the output OFDM signals satisfy the limiting threshold, the phase shift is ceased; otherwise, Q times of the cyclical phase shift is conducted, and if the peak value of the OFDM signals is still greater than the limiting threshold, the minimum signal of the peak value in the OFDM signals with the Q times of the phase shift is output.

The invention discloses a finite time output feedback attitude tracking control method for an aircraft. According to the finite time output feedback attitude tracking control method, in the attitude tracking control process of the aircraft, a tracking error of an attitude parameter vector and an expected attitude position vector of the aircraft is converged in an expected neighborhood through thesteps of: constructing a state observer, a finite time command filter and an error compensation mechanism, designing a control law based on fractional power and the like. By adopting the finite time output feedback attitude tracking control method, the problem of calculation complexity caused by a traditional backstepping method can be avoided, the attitude tracking error can be ensured to be converged into any expected neighborhood under the conditions that the aircraft has inertia uncertainty, external interference and input saturation, and all signals in a closed-loop system are bounded within limited time. In addition, the invention further provides a specific example to verify the effectiveness of the finite time output feedback attitude tracking control method, and the effectivenessof the finite time output feedback attitude tracking control method is shown through a simulation result.