350 results about "Sigmoid function" patented technology

The invention relates to a sensitive information desensitization method and system for data sharing. According to the method, by adopting statistics, a natural language processing technology and a machine learning technology, the protection of sensitive data in the whole process from data publishing to data application and usage, and automatic identification of sensitive information of such as named entities and addresses is put forward on the basis of the establishment of a sensitive information keyword library, named entities, addresses and the like is put forward; a Sigmoid function is used for calculating the correlation degree of sensitive attributes; a desensitization strategy is carried out in a combined mode of establishing a sensitive attribute generation rule library and adopting a named entity desensitization rule and a core desensitization algorithm; deep desensitization calculation is conducted in combination with numerical sensitive attributes and classified sensitive attributes separately, the desensitization degree of a whole data set is obtained, and the controlled output of data and the like are achieved by downloading a link address hash; thus, the safety of data sensitive information and a sensitive information processing strategy which meets analysis and mining requirements to the maximize degree can be ensured, and the method and system have the advantages of being good in desensitization effect, high in reliability and the like.

In layered VDR coding, inter-layer residuals are quantized by a non-linear quantizer before being coded by a subsequent encoder. Several non-linear quantizers are presented. Such non-linear quantizers may be based on sigmoid-like transfer functions, controlled by one or more free parameters that control their mid-range slope. These functions may also depend on an offset, an output range parameter, and the maximum absolute value of the input data. The quantizer parameters can time-vary and are signaled to a layered decoder. Example non-linear quantizers described herein may be based on the mu-law function, a sigmoid function, and / or a Laplacian distribution.

Systems and methods for determining characteristic transition values such as elbow values in sigmoid or growth-type curves, such as the cycle threshold (Ct) value in PCR amplification curves. A double sigmoid function with parameters determined by a Levenberg-Marquardt (LM) regression process is used to find an approximation to a curve that fits a PCR dataset. Once the parameters have been determined, the curve can be normalized using one or more of the determined parameters. Normalization is advantageous for determining the Ct value if one chooses the arbitrary fluorescence level (AFL) approach to calculating Ct values for amplification curves. After normalization, the normalized curve is processed by applying a root-finding algorithm to determine the root of the function representing the normalized curve, which root corresponds to the Ct value. The Ct value is then returned and may be displayed or otherwise used for further processing.

Systems and methods for identifying and removing spikes in data sets representing PCR growth curves or other sigmoid type curves or growth curves. A double sigmoid function with parameters determined using a Levenberg-Marquardt regression algorithm is used to find an approximation to the curve, and a statistical test such as a z-test is then used to identify spikes by identifying data points in the data set that do not fit well with the approximation. The identified spike(s) are removed from the data set and / or replaced with interpolated data points determined by using data points surrounding the identified spike(s). In one aspect, a spline interpolation process such as a cubic spline interpolation process is used to find an approximation to the data set with the identified spike points removed. Interpolated values to replace the spike points are then calculated using the cubic spline interpolation approximation curve.

The invention administers an objective clinical test to an eye that measures the visual sensitivity of the superior retina and the inferior retina, by alternately presenting a stimulus pair comprising a shaped superior light stimulus and a shaped inferior light stimulus that are horizontal mirror images of one another and have shapes encompassing visual field defects. The shaped superior and inferior light stimuli stimulate pupillary responses whose amplitudes are measured. A cycle-averaged pupillary response balance and a luminance ratio are computed for each presentation of a stimulus pair. A stimulus pair response curve is computed by fitting cycle-averaged pupillary response balances to a sigmoid function of the luminance ratios. A balanced luminance ratio at which the cycle-averaged pupillary response balance is equal to about zero is computed from the sigmoid function. The balanced luminance ratio is indicative of the presence and location of retinal nerve damage.

The invention provides a speech recognition method adopting a rectified deep auto-encoder network. The method comprises the following steps: first, carrying out training on the deep auto-encoder network by adopting a rectification linear unit as an activation function which replaces the traditional Sigmoid function, then, introducing L2 regularization for optimizing the overfitting problem easily occurring during the deep model training process, and finally, carrying out the layer-by-layer greedy and unsupervised 'pretraining' and supervised 'fine tuning', thus obtaining weights with character representation for speech recognition. According to the method provided by the invention, the strong capacity of learning substantive characteristics of data sets from the minority of samples of the deep neural network is fully utilized, thus the problems of gradient disappear and overfitting during the training process are solved, and the recognition precision of the system under the noise environment is improved.

The invention relates to a method of real-time detection for the spot welding nugget diameter of an aluminium alloy resistance, which includes the following steps:1. acquiring the electrode displacement signals in the process of spot welding and drawing the curve diagram of the electrode displacement signals;2. extracting the two eigenvalues of expansion displacement and forging displacement;3. unripping the aluminium alloy welding board, measuring the spot welding nugget diameter of the resistance and constructing a couple-sample corresponding to the extracted eigenvalues and the measured nugget diameters;4.repeating the steps 2 and 3 so as to obtaining the couple-samples with the required quantity for design; 5.constructing an artificial neural network model and carrying out training with the obtained couple-samples according to BP algorithm so as to realizing the mapping from the eigenvalue to nugget diameter, wherein the artificial neural network model has two inputs, one output and an implicit strata in the middle, which has five nodes and the transfer function of which is Sigmoid function, and the transfer function of the output layer is linear function; 6.applying the trained model to the real-time detection for the spot welding nugget diameter of an aluminium alloy resistance.

The invention discloses an oil-immersed transformer fault diagnosis method based on a fault probability distribution model, aiming to provide the probability of various faults of the oil-immersed transformer according to analysis of dissolved gas in oil. The oil-immersed transformer fault diagnosis method includes the steps of collecting massive analysis data of the dissolved gas in the oil of the transformer and normalizing the data, utilizing a Sigmoid function to realize probability output on the basis of a support vector machine, computing according to the collected massive analysis data of the dissolved gas in the oil to acquire optimal parameters of the support vector machine and the Sigmoid function, fusing the probability output of multiple support vector machines into probability output of multi-class problems, and obtaining a diagnosis result by referring to the maximum probability classification and standard deviation. The oil-immersed transformer fault diagnosis method based on the fault probability distribution model has high accuracy without error diagnosis, can be used for processing multiple fault conditions effectively and has high practicability.

An apparatus, system, and method are disclosed for predictive control of a turbocharger. The method includes interpreting a compressor performance model for a turbocharger, and interpreting at least one current operating parameter. The method further includes calculating a performance margin, calculating a performance margin derivative, and calculating a response value. The performance margin comprises a choke margin or a surge margin according to the position of an operating point in the compressor performance model. The performance margin is implemented in a first sigmoid function, and the performance margin derivative is implemented in a second sigmoid function. The response value is determined by applying a MIN function to the output of the product of the sigmoid functions in the choke margin case, and by applying a MAX function to the product of the sigmoid functions in the surge margin case.

The invention discloses a permanent magnet synchronous motor state estimation method based on a sliding mode back EMF observer. According to stator voltage and stator current detected in real time, a sigmoid function is used to substitute a sign function, and the sliding mode back EMF observer is constructed to estimate the back EMF. The type of the sliding mode back EMF observer is characterized in that the observer is insensitive to switch noise and has a filtering characteristic, thus the filtering of the estimated back EMF through a low-pass filter is not needed. After the normalization processing of the back EMF, a rotor position and speed information are extracted through a PI-I structure phase-locked loop position observer equivalent to a simplified Kalman filter. The permanent magnet synchronous motor state estimation method of the invention has good estimation performance even in the condition of noise and motor parameter changes, the chattering phenomenon in a traditional sliding mode observer is reduced, and the dynamic tracking performance and estimation precision of the rotor position and speed are improved.

Systems and methods are provided for analyzing data to determine properties of a PCR process or other process exhibiting amplification or growth. Data representing an amplification can be distinguished from data representing a jump or other error. A modified sigmoid function containing a drift term may be used in determining the properties. A multi-stage functional fit of the amplification data can provide increased accuracy and consistency of one or more of the properties. A baseline of the amplification data can be determined by analyzing an integrated area of a first derivative function of the data. A reference quantitation value can also be determined from locations of maxima of different derivative functions of the amplification data, e.g., a weighted average of the maxima locations for the second and third derivatives may be used.

The invention discloses a sigmoid function fitting hardware circuit based on a Remez approximating algorithm. A sigmoid function fitting method is characterized by including the following steps that (1), the order of a fitting polynomial is determined; (2), a fitting interval of a sigmoid function is obtained; (3), a piecewise interval is obtained; (4), the fitting polynomial is obtained; (5), a coefficient storing module is designed; (6), a polynomial calculation module is designed; (7), a judgment module is designed; (8), the fitting hardware circuit is obtained; (9), a fitting execution interval where operands are located is judged; (10), a coefficient of the fitting polynomial is read; (11), fitting calculation is carried out in the polynomial calculation module. On the basis of reducing hardware resource consumption, the calculation accuracy is improved, the calculation speed is increased, and the flexibility of a calculation structure is promoted.

The invention discloses a method for selecting an access network in a heterogeneous network. The method includes the following steps: building of a non-cooperative game model between a user and a network according to a game theory method; model solution which determines an optimizing strategy for balancing mutual interest of the network and the user; and experiment simulation and numerical analysis. A non-monotone network service quality quantization model is built by combining a gray level relevance idea and a sigmoid function in a robot study mechanism. Simulation shows that the model fits an actual heterogeneous fusion network scene, according to the non-cooperative game model between the user and the network, the strategy of balancing the mutual interest of the user and the network is adopted, and a network optimization pricing strategy is determined by solving the non-cooperative game model. In addition, the method helps the user to select the most suitable access network, accordance is provided for formulating a QoS standard of the heterogeneous network fusion system, and service requirements of different levels of the user under different scenes are met.

The invention discloses a sliding-mode variable structure control method of a variable exponential coefficient reaching law of a permanent magnet synchronous motor. The method comprises the following steps: aiming at the requirements of a control system of a high-performance permanent magnet synchronous motor, a variable exponential function term formula is introduced on the basis of a conventional exponential reaching law; the formula takes a first-order norm of systematic state variables as an exponent of exponential functions, adaptively adjusts the reaching speed of a variable exponent term and a variable speed term according to the distance between the system and a sliding-mode surface, and is helpful to increase the dynamic response speed of the system; the systematic state variables are associated by taking the first-order norm of the systematic state variables as the exponent of the functions; and an s function is introduced to replace a symbolic function for further suppressing system chattering. Therefore, the sliding-mode variable structure control method of the permanent magnet synchronous motor using the variable exponential coefficient reaching law is provided. Compared with an integral-type sliding-mode variable structure control method, the method disclosed by the invention can be used for effectively increasing the dynamic characteristics and steady state characteristics of the system as well as increasing the robustness of the system.

A method for calculating a neuron layer of a multi-layer perceptron model that includes a permanently hardwired processor core configured in hardware for calculating a permanently predefined processing algorithm in coupled functional blocks, a neuron of a neuron layer of the perceptron model being calculated with the aid of an activation function, the activation function corresponding to a simplified sigmoid function and to a simplified tank function, the activation function being formed by zero-point mirroring of the negative definition range of the exponential function.

The invention discloses a PMSM rotor position detection device and method based on an improved sliding-mode observer. A second-order generalized integration method is employed instead of a low pass filter to process a counterelectromotive force signal; an equivalent counterelectromotive force signal is observed through the improved sliding-mode observer based on a sigmoid function, buffeting of equivalent counterelectromotive force is inhibited through the second-order generalized integration method, phase lag caused by the low pass filter is avoided, a normalized quadrature phase locked loop is employed to estimate a rotor position signal according to the estimation counterelectromotive force signal, and the inhibition effect on each harmonic contained in the counterelectromotive force signal can be realized through resonance of the improved sliding-mode observer at the fundamental frequency. The method is advantaged in that the structure of the improved sliding-mode observer is simple, the observed counterelectromotive force signal is relatively accurate, and the high-precision rotor position information can be acquired.

Systems and methods for improving Ct determination in PCR amplification curves by correcting PCR data for temperature shifts that may occur during the PCR process. A double sigmoid function with parameters determined by a Levenberg-Marquardt (LM) regression process is used to find an approximation to the portion of the curve in the region after the temperature shift, termed “CAC”, the cycle where the temperature shift occurred. A robust linear approximation is determined for the portion of the curve in the region before the temperature shift. Values of the fluorescent intensity for the cycle CAC or CAC+1 are determined using both the linear approximation and the LM process, and a difference in these values is subtracted off of the portion of the data set representing the portion of the curve before the temperature shift occurred to produce a shift-corrected data set. The shift-corrected data set may be displayed or otherwise used for further processing.

The liquid crystal display device inputs a signal, corrected by a correction circuit provided with a sigmoid inverse function generator, to a source driver. A sigmoid function which is an S-shaped function is used to approximate an input / output characteristic (applied voltage-transmittance curve) of a liquid crystal panel. The sigmoid inverse function generator converts an input signal by using an inverse function of the sigmoid function obtained by performing the approximation. A γ conversion circuit provided on the correction circuit and the sigmoid inverse function generator realize the desired γ characteristic regardless of a characteristic of the liquid crystal panel. Further, since the inverse function of the sigmoid function which is an S-shaped function is used, it is possible to provide a display device which corrects the input / output characteristic of the display panel by performing the approximation in accordance with not plural functions but a single function (without dividing the characteristic into sections).

The invention relates to a detecting method of the rotor position and rotating speed of a permanent magnet synchronous motor, and belongs to the field of permanent magnet synchronous motor control. The detecting method includes the following steps that a counter electromotive force equation and a state equation of the permanent magnet synchronous motor are set up; matrix-vector arrangement is carried out on the state equation set up, and a state estimating equation is set up according to the state equation obtained after arrangement to obtain a state error equation; a second-order sliding-mode observer equation is set up according to the state error equation to obtain a counter electromotive force error equation; the sigmoid function is used for replacing the control function signum to correct the counter electromotive force error equation; a counter electromotive force estimating equation is set up according to the counter electromotive force error equation obtained after correction; the counter electromotive force estimating equation is subtracted from the counter electromotive force equation, the Lyapunov equation is used for analyzing stability to obtain and correct a rotor rotating speed estimating equation; rotor position information is extracted through the phase-locked loop technology. A model reference adaption and second-order sliding-mode combined observer is used for estimating counter electromotive force and rotor speed of the permanent magnet synchronous motor and replacing a traditional sliding-mode observer to obtain the rotor speed through counter electromotive force numerical calculation.

The invention discloses a haze weather image enhancement method. The haze weather image enhancement method comprises steps that an original haze weather color image is converted from RGB space to HSV space; filtering processing of brightness V is carried out, and color tone H, saturation S, and the brightness V are synthesized together to acquire an image 1; the image 1 is converted from the HSV space to the RGB space to acquire an RGB space image 2; the image 2 is processed by using a multi-dimensional Retinex algorithm, and the haze weather color image is divided into an incident component L and a reflection component R; the Gamma conversion of the incident component L is carried out, and the double-sided filtering of the reflection component R is carried out, and then an image 3 is acquired; the enhancement processing of the image 3 is carried out by using a Sigmoid function, and a final image 4 is acquired.

global stress constrained topology optimization method for compliant mechanisms based on K-S function is presented. The method comprises the following steps: (1) defining a design domain and initializing design variables; (2) punishing the stiffness of the structural element and obtaining the displacement and stress of the structure by the finite element method; (3) Using K-The S function condenses the local stress constraints of all elements into a global maximum stress constraint. (4) sloving Sensitivity of cross-strain energy objective function, sensitivity of volume and maximum stress constraint; 5) modifying that sensitivity of the sensitivity filtering technology; (6) Moving asymptotic algorithm is used to solve the optimization problem, and the optimal topology is obtained by iterating until the convergence criterion is satisfied. The S-function condenses the large number of constraints caused by the stresses in all elements into a global constraint, which greatly improves the computational efficiency. The topology optimization method of the invention enables the compliant mechanism configuration to avoid the hinge-like structure, meets the strength requirements, and alleviates the stress concentration phenomenon.

The invention discloses a human behavior recognition and classification method based on an accelerometer. The method includes the first step of collecting human behavior samples as a training set, the second step of searching for a projection matrix U which is optimal for the recognition and classification of the training set, the third step of carrying projection on no-labeled data, and the fourth step of classifying the projected data by using a minimum distance classifier to obtain a recognition result. According to the human behavior recognition method, a partial approximate linear hypothesis is carried out on adjacent blocks formed by labeled data so as to enable the distance between different types of samples on the blocks to be large enough, positional sequence information of the same type of samples is reserved as far as possible through class sigmoid function penalty factors, and finally a global objective function is established on the basis of the objective functions on all blocks. The human behavior recognition method can reserve the information of the distance between the samples in a higher dimensional space properly, and reduces dependence of recognition models on artificial tagging samples, and the recognition effect is superior to a representative human behavior recognition method based on linear discriminant analysis.

The invention discloses a large-scale repeated video retrieval method based on random multi-view hash. Video key frames are extracted through a time interval sampling method, and color histogram HSV and local binary pattern LBP characteristics of the key frames are extracted; the HSV and LBP characteristics are mapped to quasi-hash space through linear mapping and the sigmoid function, and hash codes representing videos finally are obtained through quasi-hash codes generated through thresholding processing; the Hamming distance between a test video hash code sequence and other video hash code sequences is calculated, and approximate repeated videos of the test video are retrieved according to the Hamming distance. Multi-view characteristic information of the videos is used, retrieval efficiency and accuracy are improved through hash mapping, and large-scale quick retrieval of the approximate repeated videos is achieved.

The invention discloses a spacecraft automatic rendezvous obstacle avoidance method and system. The method comprise the steps of solving the shortest Euler distance under the ellipsoid, constructing atarget attraction potential function, constructing an obstacle avoidance potential function under the sigmoid function, designing a controller based on a speed artificial potential function, and designing a suboptimal speed artificial function controller based on the state dependent Riccati equation. The method is applicable to spacecraft automatic rendezvous obstacle avoidance control, and can calculate the distance between the spacecraft and a target spacecraft or an obstacle more accurately through determining the shortest Euler distance under the ellipsoid, thereby being capable of more accurately controlling the trajectory of the spacecraft, and reducing unnecessary fuel consumption; and meanwhile, the sigmoid function is introduced into the aerospace field for the first time to define the obstacle avoidance potential function so as to form the overall potential function and thus design appropriate control laws, thereby being capable o more accurately controlling the trajectory of the spacecraft when being away from or getting close to the obstacle, and having more accurate flight trajectory and higher response speed.

The invention relates to a lane keeping control method. The method comprises the following steps: a camera is used for lane line detection; the distance between a vehicle and the lane line is detected; the detected distance is subjected to normalization processing; an optimized enhanced signal is designed on the basis of a Sigmoid function; and according to the numerical value of the enhanced signal and the numerical value of offset normalization processing, the system performs a corresponding control decision behavior. On one hand, a machine learning method is adopted to set continuous enhanced signals, and compared with supervised learning, the reliability, the comfort and the stability of the vehicle lane keeping function are improved; and on the other hand, compared with the prior art, the vehicle does not repeatedly swing in the lane middle line, which is closer to human driving.

The invention, which relates to the field of the spatial location and intensity information estimation of radioactive sources, provides a radioactive source positioning and intensity estimation methodin a mobile robot's nuclear environment, thereby solving problems that the mobile robot not only can not searching for the source based on the gradient information but also locate the radioactive source based on a regression algorithm. According to the invention, the involved algorithm is based on the particle filter framework, concepts of a measurement active range and a weight correction factorare introduced and a particle swarm is driven to move towards a true state based on a Poisson observation model; the particle weight is corrected by using Gaussian process regression and a sigmoid function, so that the predictive convergence of the particle group is accelerated; clustering processing is carried out on the state space by introducing a mean-lifting algorithm technology and the location of an effective radioactive source is identified and calculated rapidly. Therefore, high-precision and high-efficiency positioning of the radioactive source are realized; the robustness to the environmental radiation drift is high. The method is suitable for the radioactive source localization.