202 results about "Excitation function" patented technology

The invention discloses a face recognition method based deep learning and a face recognition device thereof and electronic equipment. The method comprises the steps that a convolutional neural network model is constructed, and the convolutional neural network model comprises a first convolution unit, a first pooling layer, multiple convolution combinations, a second pooling layer and full connection layers which are connected in turn, wherein the first convolution unit comprises a first convolution layer, a batch normalization layer and an excitation function layer, the excitation function layer simultaneously uses a ReLU function and a NReLU function to act as the excitation function, and the adjacent convolution combinations are connected by the short circuit layer of the residual network; and the convolutional neural network model is trained, the training data are inputted to the convolutional neural network model and training is performed by using the stochastic gradient descent method, and the last full connection layer is removed out of the trained convolutional neural network model and then only forward propagation is performed so as to act as the face feature data required for face recognition. ReLU + NReLU are used as the excitation function so that the computational burden can be reduced, the accuracy can be guaranteed, the model size can be reduced and the operation speed can be enhanced.

A method for the nondestructive testing of a test object by ultrasound is provided, the method including generating a pulsed ultrasonic field in the test object by means of an array of individually drivable ultrasonic transmitting transducers acoustically coupled to the test object. The ultrasonic transmitting transducers are each driven with a specific analog transient excitation signal, wherein each analog transient excitation signal is generated based on an ultrasonic transmitting transducer-specific stored digital transient excitation function. The method further includes receiving resulting echo signals from the test object by means of an array of individually drivable ultrasonic receiving transducers, with each ultrasonic receiving transducer providing an analog time-resolved echo signal, temporarily storing the time-resolved, transducer-specific, digitized echo signals in the form of an echo signal set, and applying a plurality of different reception processing rules to the echo signal set.

The invention discloses a unit, a device and a method for simulating biological neuron and neuronal synapsis on the basis of chalcogenide compounds. The unit comprises a first electrode layer, a function material layer and a second electrode layer. During the neuron simulation, a device receives the stimulation of one or a plurality of electric pulses, the resistance of the function material is changed into the low resistance state from the high resistance state, the simulated neuron is changed into an excitation state from a resting state, and the threshold value excitation and energy accumulation excitation functions are realized. During the neuronal synapsis simulation, the electric conductance of the function material layer of the device can be gradually changed according to input signals, and the synapsis weight regulating function is realized, the ynapsis weight is changed according to time differences of signals input at two ends, and the STDP (spike timing dependent plasticity) function of synapsis is realized. The basic device forming the artificial neural network can be provided.

The present invention is directed to provide a linear actuator capable of propagating magnetic force of a stator magnet within a yoke in a smooth manner so as to form a superior magnetic path. A linear actuator is provided with a moveable part facing a stator magnet provided together with a coil within a yoke, with both magnetic pole pieces of the yoke being exited to an S pole and an N pole respectively by switching of energizing of the coil so as to subject the moveable part to thrust and bring about reciprocal driving. According to the present invention, the coil is divided into split coils so as to retain excitation function, and a magnetic path body for connecting the stator magnet and the yoke is provided between the split coils. Magnetic force of the stator magnet is transmitted to the yoke via the magnetic path body.

The invention discloses a double-stator disc type mixed excitation motor which comprises two stators, a permanent magnet, a three-phase concentrated armature winding, a single-phase concentrated excitation winding and a rotor. The stators and the rotor are arranged coaxially, and the rotor is arranged between the two stators. The permanent magnet, the armature winding and the electric excitation winding are all located on the stators. The rotor is free of the permanent magnet or the windings and simple in structure. An air-gap magnetic field is formed by an electric excitation magnetic field generated by the electric excitation winding and a permanent magnetic field generated by the permanent magnet, and adjusting of the air-gap magnetic field is achieved by changing electric excitation current. The motor structurally reserves the characteristics of a disc type magnetic flux switching permanent magnetic motor of being compact, simple and suitable for high-speed operation, can achieve a mixed excitation function without increasing the motor size additionally and guarantees good torque output capacity and high power density. The motor is small in axial size and suitable for strict and thin type installing in places of a hub driving motor for electromobile and the like.

The invention relates to a distance sense synthetic method in three-dimensional sound field synthesis. The method comprises firstly setting reverberation time of a room to be synthesized and the length, the width and the height of the room; or using acoustic pressure levels to measure reverberation time of a room to be replayed and the length, the width and the height of the room; secondly measuring frequency response of a loudspeaker, obtaining a reflection coefficient beta i according to a formula; and finally obtaining corresponding excitation functions of each loudspeaker according to setting position coordinates of each loudspeaker and each sound source. In the formula, P (chi, omega) is sound pressure synthesized in a bounded area V, and Ssw(omega) is a source signal frequency domain expression. Therefore, relation between room acoustics and traditional WFS technology is built, and the excitation function is finally used to enable a three-dimensional sound field to have direction sense and the distance sense.

A light mixing type monochromatic light source with the function of fluorescent powder excitation function comprises a first pumping light source module for emerging first pumping light, a second pumping light source module for emerging second pumping light, a color combination mirror, a light filter, a fluorescent powder layer and an optical lens, wherein the light filter, the fluorescent powder layer and the optical lens are arranged in sequence. The fluorescent powder layer is used for absorbing the first pumping light and the second pumping light and exciting the fluorescent light whose wavelength is different from that of the first pumping light and that of the second pumping light. A first thin film is coated on the surface of the optical lens facing to the fluorescent powder layer and is used for transmitting the fluorescent light and reflecting the first pumping light and the second pumping light. A second thin film is coated on the surface of the light filter facing to the fluorescent powder layer and is used for transmitting the first pumping light, the second pumping light and reflecting the fluorescent light. The monochromatic light source increases the excitation energy of fluorescent powder in a manner of mixing the two pumping lights, thereby increasing the emergence brightness of the fluorescent light. Compared with the traditional large chip, the light mixing type monochromatic light source has low cost and increases the power of finally-emerged monochromatic light. In addition, the invention also provides a projection optical engine using the light mixing type monochromatic light source.

The invention discloses a dynamic gesture learning and identifying method based on a Chebyshev neural network. Chebyshev orthogonal polynomials serve as hidden-layer neuron excitation functions for constructing a multi-input multi-output three-layer feedforward neural network, and a weights direct determination method and a hidden-layer node number adaptive determination algorithm are given; a fingertip detection algorithm based on a color histogram and a fingertip tracking algorithm based on bigraph optimal matching are given for obtaining a dynamic gesture track in real time; an MIMO-CNN (multi-input multi-output Chebyshev neural network) is subjected to input output structure design and network weights learning training according to the dynamic gesture identifying requirements, and a dynamic gesture is identified by the trained MIMO-CNN. A test result shows that the MIMO-CNN can increase the network training speed and improve the network training precision, so that the dynamic gesture learning speed is increased and the dynamic gesture identifying accuracy is improved; moreover, relatively good robustness and generalization ability in the aspect of dynamic gesture identification are achieved.

The invention discloses an excitation structure and an alternating-current and direct-current excitation control method for a frequency-variable alternating-current starting power generation system. The excitation structure consists of an open circuit-type stator three-phase winding and a three-phase bridge-type converter, wherein the two ends of the three-phase bridge-type converter are connected. The method comprises the following steps of: starting an initializing process; coordinately controlling the output vector of each converter; selecting a proper comprehensive voltage vector; generating a required three-phase alternating-current excitation voltage in the three-phase winding so that a magnetizing exciter works in a rated output state to provide enough excitation current for the starting and the operation of a main power generator; when the rotating speed of a motor reaches an excitation switching point, switching from a three-phase alternating-current excitation mode to direct-current excitation according to a rotating speed feedback signal, wherein the structure of the winding of the magnetizing exciter does not need to be changed by an external switch in the switching process; and making the three-phase alternating-current winding equivalent to a direct-current serial structure by directly controlling each phase of winding current respectively by utilizing the bridge arms of two converters. Therefore, a direct-current excitation function is realized.

The invention relates to a passenger flow prediction method based on a time-lag NARX neural network and mainly aims to solve the technical problem of low prediction precision in the prior art. The method comprises the steps that n pieces of historical data is collected from an automatic fare collection system to serve as original samples, and preprocessing is performed to obtain preprocessed samples; an NARX short-time passenger flow prediction model p(t) about a time sequence is established according to a nonlinear autoregression network with external input, wherein the external input is an external impression factor u(t), and the nonlinear autoregression network with the external input comprises an input layer, input lag time, a hidden layer, an output layer and output lag time; and real-time passenger flow prediction is performed according to the NARX short-time passenger flow prediction model, an excitation function and a training algorithm, wherein real-time passenger flow prediction comprises short-time passenger flow prediction, peak prediction and representative passenger flow distribution site prediction. Through the technical scheme, the technical problem is solved, and the method can be applied to rail passenger flow prediction.

The invention discloses a detection device, method and system for internal defects of a ceramic insulator. The method comprises the following steps: applying a light pulse heat excitation function toa ceramic insulator to be tested by using a light pulse heat excitation loading device; acquiring infrared thermal image sequences on the surface of the ceramic insulator to be tested by using an infrared thermal image acquisition device, and carrying out preprocessing such as fitting, compression and reconstruction on the acquired infrared thermal image sequences by using a data processor; according to defect information provided by original infrared heat image sequences obtained through preprocessing, analyzing and comparing data in a characteristic database, so as to obtain defect types andquantitative calculation results of the ceramic insulator. By adopting the detection device, the light pulse heat excitation function is applied to the surface of the ceramic insulator, defect typesare judged and defect information is analyzed according to variations of dynamic temperature fields on the surface of the ceramic insulator, the detection device has the characteristics of rapid measurement, large single-time measurement areas, visible measurement results, non contact, and the like, and related information of internal defects of the ceramic insulator can be accurately, comprehensively and deeply processed and analyzed.

The present invention is a system and method for characterizing human (or animate) speech voiced excitation functions and acoustic signals, for removing unwanted acoustic noise which often occurs when a speaker uses a microphone in common environments, and for synthesizing personalized or modified human (or other animate) speech upon command from a controller. A low power EM sensor is used to detect the motions of windpipe tissues in the glottal region of the human speech system before, during, and after voiced speech is produced by a user. From these tissue motion measurements, a voiced excitation function can be derived. Further, the excitation function provides speech production information to enhance noise removal from human speech and it enables accurate transfer functions of speech to be obtained. Previously stored excitation and transfer functions can be used for synthesizing personalized or modified human speech. Configurations of EM sensor and acoustic microphone systems are described to enhance noise cancellation and to enable multiple articulator measurements.

The invention provides a prediction method and system of internal combustion gas concentration LSTM of a comprehensive pipe rack. The prediction method of internal combustion gas concentration LSTM ofa comprehensive pipe rack includes the steps: acquiring the internal combustion gas concentration data of the comprehensive pipe rack, the internal microbe density data of the pipe rack, the sensor position data and the positional data of the comprehensive pipe rack; preprocessing the acquired internal combustion gas concentration data of the comprehensive pipe rack, the internal microbe densitydata of the pipe rack, the sensor position data and the positional data of the comprehensive pipe rack, and forming combustion gas concentration characteristic vectors; taking ReLu as an excitation function to train an LSTM model; and inputting the combustion gas concentration characteristic vectors into the LSTM model, and correspondingly outputting the predicted data of the internal combustion gas concentration of the comprehensive pipe rack. The prediction method and system of internal combustion gas concentration LSTM of a comprehensive pipe rack can construct a model by means of a recurrent neural network with a long short term memory, and is adapted to dynamic characteristics of the internal combustion gas concentration data of the comprehensive pipe rack so as to realize long-term accurate prediction of the internal combustion gas concentration data of the comprehensive pipe rack in advance. The prediction method and system of internal combustion gas concentration LSTM of a comprehensive pipe rack has the advantages of high universality, low cost and wide applicability.

The invention discloses a high-consistency photoacoustic excitation and detecting integrated fiber probe, a manufacturing method and a testing method thereof. According to the probe, a light absorbingmaterial (such as graphene or molybdenum disulfide and the like) and an elastic material (PDMS) are combined and deposited on the end surface of a double-clad fiber, and high-power pulse laser is utilized for performing photoacoustic signal excitation. Furthermore a light-absorbing sensitive material and a single-mode fiber end surface form a Fabry-Perot (F-P) interference chamber for realizing photoinduced ultrasonic signal detection. The probe has a photoacoustic excitation function and a detecting function and can be manufactured and formed in one time. Furthermore because the light absorbing material is prepared through a chemical liquid phase deposition method, the high-consistency photoacoustic excitation and detecting integrated fiber probe has advantages of controllable thickness,low number of steps, high repeatability, etc. Therefore, the fiber type photoactoustic probe has advantages of photoacoustic signal receiving-and-transmitting integration, small size, simple manufacture, high consistency, high photoacoustic conversion efficiency, high electromagnetic interference resistance, etc. The high-consistency photoacoustic excitation and detecting integrated fiber probe can be used in the fields such as industrial nondestructive detection, biomedical ultrasonic and material characteristic measurement.

The embodiment of the invention provides a call stack analyzing and processing method and device. The method comprises the following steps: searching layer by layer according to a preset byte number, judging if the search value of a current position is a valid return address value, if yes, then obtaining the valid return address value and the stack pointer of the current position; obtaining the size of the open stack and the storage offset of the valid return address value in the layer; connecting each valid return address value in series according to the corresponding stack pointer and the size of the open stack and the storage offset in the layer, and obtaining a candidate call stack; and screening out an objective call stack from at least one candidate call stack according to a preset screening rule. According to the embodiment of the invention, a plurality of valid RA (return address) values are obtained, the valid RA values and the excitation functions of the valid RA values are judged, and then the valid RA values are connected in series to get at least one candidate call stack from which the objective call stack is screened out, so that a preferred call stack analysis result is obtained.

The invention relates to an extreme learning machine classifying method based on waveform addition Cuckoo optimization. The extreme learning machine classifying method mainly comprises the steps that (I) a training sample matrix is established; (II) M initial parasitic nests are generated on each hidden node; (II) the classifying accuracy of a waveform addition extreme learning machine classifying model is solved; (IV) training samples are randomly and equally divided into parts (please see the number of the parts in the specification), and the classifying accuracy output value of the extreme learning machine classifying model verified in a cross mode is solved; (V) an inverse hyperbolic sine function and a Morlet wavelet function are superposed to serve as an excitation function of the extreme learning machine, the waveform addition extreme learning machine classifying model is structured, and the current generation classifying accuracy of a Cuckoo algorithm is obtained; (VI) a next generation result of the Cuckoo algorithm is solved, and parasitic nests are newly established with the probability Pa; (VII) repeated iteration is conducted, whether the iteration is ended is judged, an optimal extreme learning machine classifying model is established if ending conditions are met, and the optical extreme learning machine classifying model is used for classifying unknown samples. The extreme learning machine classifying method is low in calculation complexity, high in efficiency, stable in classifying performance, high in accuracy and high in global optimization and generalization performance.

The invention discloses a feature-guided-wave-based weld seam defect detection dedicated array type sensor time-sharing excitation system and method. The feature-guided-wave-based weld seam defect detection dedicated array type sensor time-sharing excitation system comprises a master controller, a digital analog converter, a signal conditioning module, an analog multiplexer module, a phase shiftergroup and a sensor. According to the feature-guided-wave-based weld seam defect detection dedicated array type sensor time-sharing excitation method, under the circumstance that a single controller generates a one-way signal, a time-sharing excitation function of an array type sensor can be realized to enable signals excited by a single crystal oscillator in the array sensor to aggregate into anultrasonic guided-wave signal which has a high signal to noise ratio and high energy and can propagate remotely in a weld seam. The feature-guided-wave-based weld seam defect detection dedicated arraytype sensor time-sharing excitation system can be applied to weld seam defect detection of a large workpiece with uneven surface curvature and a complex structure, and has wide application prospects.

The invention discloses a deep extreme learning machine-based hazard source identification method. A deep neural network adopted by the method consists of two parts: a deep structure module and a single-hidden layer neural network module. The method comprises the following steps of: dividing hazard source information into different domain classifications by utilizing SVM and inputting the different domain classifications into corresponding network modules; carrying out an S-ELM algorithm on each network module to obtain each pre-identification result of the deep network; combining the pre-identification results of the deep network to serve as an input of a top neural network; calculating an initial hidden layer output and an output weight of a single-hidden layer ELM according to an ELM algorithm and an excitation function; determining a final input weight, a hidden layer feature space and an output weight of the network according to an improved counter propagation algorithm; and finally obtaining a hazard source identification result. The identification method disclosed by the invention can be used for improving the flexibility of the hazard source identification, decreasing the rapid expansion of empirical data, improving the utilization rate of experiential knowledge and relieving the internal memory pressure during high-dimensional data training.

Three ends of input and three ends of output in three phases AC-AC type matrix transform unit (MTU) of the controller are connected to power supply and AC motor respectively. MTU is composed of nine pieces of two-way switch arranged in 3*3 matrix. Each input and output phase is only connected to one piece of two-way switch. In overvoltage absorption circuit, output ends in phase in two sets of rectification bridge (RB) are connected, and clamping capacitor is connected between positive and negative poles. Leak resistance is connected to clamping capacitor in parallel. Three inputs and three outputs of RB are connected to inputs and outputs of MTU. DC exciting circuit fetches out voltage at two ends of the clamping capacitor. The invention realizes AC-AC direct conversion, and bidirectional energy flows realizes four-quadrant operation of motor. Using DC in overvoltage absorption circuit, exciting circuit reduces size and weight of controller.

The invention discloses an electric power system harmonics analysis method based on a multilayered feedforward neural network in the electric power system signal testing technology field. In the invention, an electric power system voltage or current signal is obtained through an optical fiber voltage sensor or an optical fiber current sensor, single hidden layer of the multilayered feedforward neural network is established; an excitation function is a sine and a cosine function and a variable parameter is a harmonic amplitude and an angular frequency; a Hanning window is performed to an obtained electric power system signal; and then discrete fourier transform (DFT) is performed; the sine and cosine component amplitude of each corrected subharmonic and the harmonic angular frequency are taken as an initial weight value of the neural network; an RPROP algorithm training is employed on the basis of the initial weight value; the amplitude and the frequency of the each subharmonic can be acquired according to the trained weight value. By using the method of the invention, accuracy of a calculating result is high and a speed is fast. The harmonic wave analysis accuracy at short sampling time can be greatly raised. A principle is simple and the method is easy to be realized.