44 results about "Pulse computation" patented technology

The invention belongs to the technical field of machine state prediction, and particularly relates to a fault diagnosis method and system based on a coding and decoding attention mechanism enhanced pulse neural network, and the method comprises the steps: collecting a device vibration signal, and segmenting the device vibration signal into time sequence segments; converting each fragment into a two-dimensional time-frequency image by using S transformation to construct a data set; the image is input into a lightweight hybrid network model to extract deep features, the model is composed of a pulse convolution encoder and a pulse efficient additive attention mechanism encoder which are alternately cascaded, and the pulse convolution encoder extracts local features through pulse neurons and depth separable convolution; the global context dependency is modeled by adopting an efficient additive attention mechanism with low linear complexity; and finally, fault classification is completed based on deep features. Through fusion of pulse calculation and an efficient attention mechanism, the feature extraction capability and diagnosis precision are improved while the parameter quantity of the model is remarkably reduced, and the method is particularly suitable for being deployed on edge equipment with limited resources to realize fault diagnosis.

The utility model belongs to the field of distributed optical fiber sensing, and particularly relates to a compact high-signal-to-noise-ratio dark pulse BOTDR (Brillouin Optical Time Domain Reflectometer) system for engineering, which comprises a laser, and the output end of the laser is connected with a first photoelectric coupler which is used for dividing light emitted by the laser into upper light and lower light. The dark pulse modulation system module is arranged, descending pump light is equivalent to a dark pulse through modulation, a computer controls dark pulse time slots so as to change pump pulse characteristics, a corresponding demodulation method is adopted in rear-end data processing, Brillouin frequency shift along a sensing optical fiber is obtained, and the Brillouin frequency shift is converted into Brillouin frequency shift along the sensing optical fiber. The lower limit of the spatial resolution of the method is equal to the fall time of the modulation system, and the spatial resolution can be greatly improved under the condition of improving the signal-to-noise ratio. In addition, due to narrow pulse width and low energy of the traditional bright pulse, an erbium-doped fiber amplifier (EDFA) is generally needed for signal amplification. Dark pulses are close to continuous light in most cases, and EDFA is not needed, so that the system design is simplified, and the cost is reduced.

The invention discloses a time service synchronization method, a computer readable storage medium and electronic equipment, and the method comprises the steps: starting a high-frequency SPXO clock source in each preset period based on the timing of a low-frequency SPXO clock source, and enabling the high-frequency SPXO clock source to be aligned with a low-frequency clock moment after buffering; calculating the planned output time of the pulse time service signal by combining the high-frequency clock receiving window, the synchronous packet sending duration, the preset ToF correction window timing value and the latest ToF value; the synchronization packet sent by the host is received before the plan time, correction output time is determined according to the receiving moment, the ToF correction window timing value and the ToF value, and the host sends a pulse to the slave after sending the synchronization packet and passing through the ToF correction window; calculating correction and plan time deviation; performing clock taming on a low-frequency SPXO clock source and a high-frequency SPXO clock source through a multi-cycle deviation value to realize time service synchronization of the slave and the host; and then the high-frequency SPXO clock source is closed.

The invention belongs to the technical field of artificial intelligence neural networks, and particularly relates to a pure pulse-driven Transform neural network automatic architecture search method. A TransSpikingform network structure is combined, a pulse word segmentation device, a pulse Transform block and a classification head are included, and efficient feature extraction is achieved through a self-attention mechanism and pulse neuron characteristics. The invention further provides a method based on non-training neural architecture search, an optimal model is quickly screened by defining a search space and generating a candidate architecture and taking FLOPs as an index, and an evolutionary algorithm is combined for iterative optimization, so that the method adapts to different resource constraints. The method shows higher accuracy and remarkable energy consumption advantages on different data sets, can automatically search for an optimal model structure according to resource constraints, and is suitable for various equipment resource conditions. Through an innovative SN-ConvBN pulse residual connection structure, non-pulse calculation in traditional residual connection is avoided, it is ensured that all operations conform to the event-driven principle, and energy consumption is remarkably reduced.

The application belongs to the technical field of photon imaging detection, and particularly relates to a photon counting imaging detection system based on a mixed anode. The system comprises a vacuum sealed tube detector composed of a window coated with a photocathode, a microchannel plate stack and a mixed anode, and a signal processing circuit. The photocathode is used for converting photons incident through the window into photoelectrons. The microchannel plate stack is used for converting the photoelectrons into electron clouds. The mixed anode comprises a strip anode and a delay line anode arranged orthogonally, and the mixed anode transmits the collected charge short pulses of each channel to the signal processing circuit. The signal processing circuit calculates the two-dimensional coordinates of the photons based on the collected charge short pulses of each channel, and then transmits the two-dimensional coordinates of the photons to a terminal for imaging display. The application provides a photon counting imaging detection system based on a cross strip-delay line mixed anode, and solves the problem that the existing position sensitive anode detector cannot simultaneously have high spatial resolution and high temporal resolution.

The present invention belongs to the technical field related to communication engineering, and discloses a mobile Loran positioning system and its transmission signal modulation method and reception signal analysis method. The analysis method includes: obtaining the modulated pulse groups transmitted in sequence by three signal transmitting devices received by the receiving port, the three modulated pulse groups carrying the coordinate information of the main station, auxiliary station 1 and auxiliary station 2; obtaining the coordinate information of each signal transmitting device through demodulation and decoding, and restoring each modulated pulse to the reference pulse position, and then accumulating to obtain the accumulated pulse of each signal transmitting device; and calculating the receiving position based on the accumulated pulse. Through the above method, the present invention can obtain the coordinate information of the transmitting station on the one hand, support the mobile deployment of the transmitting station, and by restoring the modulated pulse to the reference pulse position and then accumulating, this method eliminates the influence of the pulse modulation position and can obtain the highest accumulated pulse signal-to-noise ratio, thereby improving positioning accuracy.

The application discloses a kind of based on dynamic pulse discrimination steam turbine speed anti-interference measurement method, it is related to industrial automation control field, the method includes: S1: by the speed sensor of installation in steam turbine rotor speed measuring gear flank alternately changed and generated sine wave signal of tooth valley conversion into digital pulse signal by conditioning circuit;S2: calculate the effective pulse frequency f of this sampling time, to calculate effective pulse cut-off period T;S3: for subsequent new pulse, calculate the time interval Δt between it and the last pulse being judged as effective;If Δt≥T, then determine that the new pulse is effective pulse;If Δt < T, then determine that the new pulse is interference pulse and is rejected;S4: update effective pulse frequency f, calculate the real-time speed of steam turbine, calculate the real-time speed of steam turbine solves the defect that present art relies on hardware filtering and is difficult to eliminate singular interference pulse completely, leading to unreliable speed measurement.

The invention discloses a reconfigurable hybrid neural network calculation method and device, and belongs to the technical field of neural networks, the reconfigurable hybrid neural network calculation device comprises a plurality of pulse calculation units which are arranged in sequence, the first input of each pulse calculation unit comprises one of first data and second data, and the second input of each pulse calculation unit comprises one of second data; the first data is one bit data of the to-be-convolved data, the second data is pulse data of a time step length, and the output film voltage of the adjacent previous pulse calculation unit is used as the second input of the adjacent next pulse calculation unit; each pulse calculation unit responds to all current input first data to obtain a current reasoning result and an output membrane voltage; the multiply-accumulate unit responds to the second data of the current time step length, the multiply-accumulate unit and each pulse calculation unit jointly obtain the output membrane voltage and the output pulse based on the first input and the second input, calculation of the pulse neural network and the artificial neural network is supported at the same time, and extra hardware resources are not consumed.

The invention belongs to the related technical field of integrated circuits, and discloses a hardware system for deploying a sparse pulse Transform model, which comprises an event input module, a time-space pruning module, a scheduling module, a sparse pulse calculation module and a classification head output module, the time-space pruning module judges the importance of each time step and each token according to a pulse accumulated value and generates a time-space pruning mask, the scheduling module screens a token sequence according to the mask to form a sparse token sequence, the sparse pulse calculation module performs sparse reasoning on the sparse token sequence, and finally, the classification head output module outputs a classification result. The system can be used for deploying a sparse pulse Transform model through a special hardware control and scheduling mechanism, and provides an accurate execution path and efficient scheduling support for dynamically changing sparse data streams.

The application provides a pulse remote signaling acquisition method and a pulse remote signaling acquisition terminal of an electric energy meter, and relates to the technical field of data acquisition. The method comprises the following steps: applying corresponding target storage resources for a pulse input port in a storage module; performing state scanning on the pulse input port to obtain current pulse output data of the electric energy meter; performing pulse calculation according to pulse configuration information of a pre-designed measurement point and the current pulse output data to obtain current pulse measurement data; acquiring current remote signaling input signals of the electric energy meter collected by a remote signaling port; determining current operating states of each controlled switch in the electric energy meter according to the current remote signaling input signals; and uploading the current pulse output data, the current pulse measurement data and the current operating states stored in the target storage resources to a master station system by using a remote communication module. The application can upload all kinds of data of the electric energy meter to the master station system through the pulse remote signaling acquisition terminal, and realizes real-time monitoring of the state of the electric energy meter.

The invention discloses a hardware accelerator supporting pulse Transform calculation and an acceleration method of the hardware accelerator. The accelerator comprises a control module, a PE calculation unit array, a matrix processing array, a full-connection core and a storage module. The PE calculation unit array is used for realizing tasks of pulse convolution calculation, pulse distribution and membrane potential maximum pooling calculation; the matrix processing array is used for realizing an operation task of a pulse matrix based on a bit AND and bit OR operation mode; and the full-connection core is used for realizing a pulse full-connection calculation task. According to the invention, acceleration support of a pulse self-attention mechanism and a Transform architecture is realized through the PE calculation unit array, the matrix processing array and the full-connection kernel, and meanwhile, the calculation energy consumption of the accelerator is reduced by adopting a bit AND and bit OR operation mode, so that the deployment universality of the accelerator is improved.

This state change detection system comprises: a rotation detection unit that outputs a pulse signal; a pulse calculation unit that obtains a pulse calculation waveform representing a pulse period for each rotation angle of a rotating body; a reference generation unit that generates a reference waveform in which an average pulse calculation waveform, which is obtained by averaging values of the pulse calculation waveform with an accumulated value for each rotation angle, is represented by the distribution for one rotation of the rotating body; a rotational fluctuation waveform generation unit that uses the reference waveform to generate a rotational fluctuation waveform obtained by extracting, from the pulse calculation waveform used to generate the reference waveform or a pulse calculation waveform succeeding the pulse calculation waveform, a fluctuation component other than a steady component that steadily occurs with the rotation of the rotating body; and a state change detection unit that detects a change in rotation state on the basis of the rotational fluctuation waveform.

The invention provides an alternating-current dynamic traceability system and method for a novel electric power system, and aims to realize analog output of various types of wide dynamic characteristics and variable power signals and test and calibrate output signals. The system comprises a man-machine interaction unit used for carrying out parameter setting on a three-phase AC dynamic power output unit; the high-precision clock unit is used for providing a clock signal and a three-phase phase synchronization control signal for the three-phase alternating-current dynamic power output unit; the three-phase alternating-current dynamic power output unit is used for setting the parameters, the clock signal and the three-phase phase synchronization control signal, generating a standard pulse and a digital waveform signal, and generating a three-phase alternating-current dynamic power signal based on the digital waveform signal; the detected electric energy meter is used for receiving the three-phase alternating current dynamic power signal, measuring the three-phase alternating current dynamic power signal and generating a corresponding detected pulse; and the error calculation unit is used for receiving the standard pulse and the detected pulse and calculating the error of the detected pulse based on the standard pulse.

The invention relates to the technical field of durability testing, in particular to a durability testing method for a new energy automobile glass lifter, which comprises the following steps of: acquiring a ratio of a current change rate to a sliding speed, judging a trend by differentiating a sliding window, generating a trend instruction, calculating current peak offset, and generating delay pulses by fuzzy classification. And acquiring symmetry offset and screening abnormal sequences, outputting an early warning level through a two-parameter clustering model, and correcting an abnormal label level through a Drools engine. According to the method, the ratio of the motor load current change rate to the glass sliding speed is dynamically monitored, the ratio of the current change rate to the sliding speed is monitored, abnormal fluctuation is analyzed and captured through a sliding window, a fuzzy logic classifier is combined to classify the current peak offset to generate delay pulses, the symmetry offset is calculated, and the trend is evaluated. The section positions are compared, the grade labels are output, multiple algorithms are fused to form closed-loop feedback, and the early warning precision and the testing efficiency are improved.

The invention relates to a method for performing a magnetic resonance measurement comprising multiple partial measurements, a magnetic resonance device, and a computer program product. According to the method, a first partial measurement is performed during a first period of time, and a second partial measurement is performed after the first partial measurement. The second partial measurement comprises the application of an excitation pulse to excite a magnetic resonance signal. During the first period of time, at least part of a pulse calculation of the excitation pulse is performed.

The present invention discloses a high-speed single-photon detection system and method. The system includes: a microchannel plate for converting incident high-speed single photons into electron clouds; an anode for dividing the electron clouds into multiple paths, generating short current pulses from the charge collected in each path and transmitting them to a signal processing path for the corresponding path; the signal processing path for converting the short current pulses into digital pulses, selecting a corresponding target shaping filter based on the time interval between adjacent digital pulses in the same path for adaptive shaping filtering, calculating the relative two-dimensional coordinate value of the high-speed single photon based on the shaped digital pulses, and transmitting the relative two-dimensional coordinate value to a terminal for imaging display. The signal processing path includes multiple shaping filters with different filtering parameters, each corresponding to a preset time interval range. By adaptively selecting the shaping filter based on the dynamic change of the pulse time interval, the present invention improves the actual count rate and spatial resolution during detection.

The application belongs to the technical field of integrated circuits, and discloses a hardware system for deploying a sparse pulse Transformer model, which comprises an event input module, a space-time pruning module, a scheduling module, a sparse pulse calculation module and a classification head output module. The event input module converts a time sequence pulse data stream into a pulse token sequence. The space-time pruning module judges the importance of each time step and each token through a pulse accumulation value, generates a space-time pruning mask, and the scheduling module screens the token sequence according to the mask to form a sparse token sequence. The sparse pulse calculation module performs sparse inference on the sparse token sequence, and finally the classification head output module outputs a classification result. The system can be used for deploying a sparse pulse Transformer model through a special hardware control and scheduling mechanism, and provides accurate execution paths and efficient scheduling support for dynamically changing sparse data streams.

The application belongs to the technical field of incremental learning, and discloses a pulse neural network model construction method for task incremental learning. A gated parameter neuron layer of the pulse neural network comprises: a synaptic current calculation unit for calculating synaptic current of an input neuron at time t based on screened pulses fired by each neuron of a previous layer; a screened mask generation unit for randomly generating a screened mask for each task; a screened membrane voltage calculation unit for multiplying the screened mask and the membrane voltage v t to obtain screened membrane voltage at time t; a pulse firing module for calculating a hidden state membrane voltage based on the obtained gating value and the screened membrane voltage to determine whether a neuron fires a pulse s t ; and a screened pulse calculation unit for calculating the fired screened pulse and transmitting the fired screened pulse to each neuron of a next layer. The application can alleviate the catastrophic forgetting problem of the pulse neural network in task incremental learning, thereby improving the classification accuracy of the model.

This invention relates to the field of laser detection and navigation positioning technology, and discloses a real-time noise reduction and in-situ calibration system and method for laser ranging sensors. The system includes an acquisition module, an extraction unit, a mapping unit, a control unit, and a compensation unit. The system extracts the zero-position calibration pulse generated by reflection within the optical window, calculates the waveform asymmetry parameter to construct a timing compensation operator for the receiving link, and the control unit adjusts the measurement echo detection threshold according to the pulse energy distribution parameter. The compensation unit uses the operator to correct the target echo phase centroid and uses the real-time phase center of the pulse as the timing start anchor point. This invention converts the ranging reference from static setting to dynamic physical reconstruction, and synchronously eliminates clock temperature drift and circuit delay deviation caused by ambient temperature rise during the ranging cycle, thereby eliminating the cumulative error caused by spatiotemporal reference drift in long-distance navigation.

An image sensor assembly includes an image sensor array having a pixel kernel. The pixel kernel includes pixel circuits, wherein each pixel circuit generates a pixel signal pulse. The pixel signal pulse has a signal pulse width, wherein the signal pulse width is a function of radiation intensity incident on the pixel circuit. The image sensor assembly further includes a plurality of analog compute modules, wherein each analog compute module receives the pixel signal pulse from a first one of the pixel circuits of the pixel kernel, outputs weighted pixel signal pulses to each of the other pixel circuits of the pixel kernel with separate weighting coefficients, receives the weighted pixel signal pulses from the other analog compute modules assigned to the pixel kernel, and generates a computed pulse having a computed pulse width. The computed pulse width is a function of the signal pulse widths of the received weighted pixel signal pulses and the weighting coefficients applied to the received weighted pixel signal pulses.

The invention discloses a pulse neural network circuit and an electronic chip, and the circuit comprises a feature storage module which is used for storing input and output pulses; the pulse extraction module is used for detecting input features according to the current calculation feature range and extracting pulses with non-zero features; the calculation engine is used for triggering and reading a local weight according to the input pulses, if all the pulses in the current calculation characteristic range are input, multiply-accumulate operation is executed, and the local weight is read into the in-memory calculation array at a time; the nonlinear module is used for executing LIF operation on the multiply-accumulate result; the pulse writing module is used for writing a result of the nonlinear module into the feature storage module for next-layer operation; wherein pulse calculation of all time steps of the current feature is completed in a time step priority mode in pulse reading, and then reading of other input features is executed. The power consumption can be reduced, and meanwhile data carrying is reduced.

Techniques are disclosed for operating a time-of-flight (TOF) sensor. The TOF may be operated in a low power mode by repeatedly performing a low power mode sequence, which may include performing a depth frame by emitting light pulses, detecting reflected light pulses, and computing a depth map based on the detected reflected light pulses. Performing the low power mode sequence may also include performing an amplitude frame at least one time by emitting a light pulse, detecting a reflected light pulse, and computing an amplitude map based on the detected reflected light pulse. In response to determining that an activation condition is satisfied, the TOF may be switched to operate in a high accuracy mode by repeatedly performing a high accuracy mode sequence, which may include performing the depth frame multiple times.

The application relates to a vehicle speed calculation method, system, calculation device and medium, the method comprising collecting a plurality of pulses generated by tire rotation once per preset timestamp, and determining the start time and end time of each preset timestamp; when no pulse generated by tire rotation is collected for a continuous preset duration, recording the current time; determining a target timestamp according to the current time; determining a statistical start time and a statistical end time according to the target timestamp; determining the first number of pulses between the start time and the end time in the target timestamp, and determining the second number of pulses between the statistical start time and the statistical end time; and calculating the vehicle speed according to the average of the first number and the second number. The application solves the problem that a complete tooth pulse cannot be collected through periodic sampling at low speed, so that the effective number of pulses collected by periodic sampling is not necessarily accurate, and the accuracy of the vehicle speed calculated according to the pulses in the period is low.

The Hall magnetic grid linear displacement sensor comprises a shell, a guide rod, a magnetic grid sliding block and a circuit board, the magnetic grid sliding block is arranged on the shell in a sliding mode, the guide rod is connected with the magnetic grid sliding block, a plurality of N poles and S poles which are distributed at equal intervals are recorded on the magnetic grid sliding block, and the circuit board is fixedly arranged on the shell. A plurality of Hall sensor chips are uniformly arranged on the circuit board at intervals along a first direction; and the guide rod can push the magnetic grid slide block to move above the Hall sensor chips along the first direction. The measurement precision improving method comprises the steps that a guide rod pushes a magnetic grid sliding block to move above a Hall sensor chip in a first direction, so that the Hall sensor chip generates periodic sine wave voltage signals or cosine wave voltage signals; the MCU converts the periodic sine wave voltage signal or cosine wave voltage signal into two orthogonal square wave signals A and B; and the MCU accumulates pulses of the square wave signals A and B to calculate the displacement variation of the magnetic grid sliding block. The detection precision is improved.

The application generally relates to a computing system including at least one processor. The at least one processor is configured to execute instructions stored in at least one memory to: initiate emission of a light pulse by an illuminator, after a predetermined delay from emission of the light pulse, initiate capturing of a plurality of pixels in a scene using a plurality of sensors based on the plurality of captured pixels, for a point in the scene, compute a respective value for volumetric density, normal, reflectance and ambient light using a corresponding neural field. The processor further, based upon the emitted light pulse, computes a shadow component corresponding to an origin of the illuminator and a direction of the emitted light pulse and using the computed respective value for volumetric density, normal, reflectance and ambient light and the computed shadow component, constructs a gated image through a volume rendering formulation.

The invention discloses a single-point real-time pulse dominant wave interval sequence extraction method and device, and relates to the technical field of two-dimensional pulse signal analysis. In order to solve the technical defects that in the prior art, challenges exist in the aspects of real-time extraction and accurate detection of pulse dominant wave interval sequences, and particularly, the application in wearable equipment and mobile terminal environments is obviously limited, the technical scheme provided by the invention comprises the following steps: acquiring a two-dimensional pulse signal to be processed; performing multi-byte data analysis on the two-dimensional pulse signal to obtain a data stream; noise reduction and baseline correction are conducted on the pulse signals, and standard pulse waves are output; converting the waveform into a square wave to obtain a high-low level signal; filtering noise pulses with insufficient high-level width; and calculating a time difference between adjacent effective pulse periods, and outputting a pulse dominant wave interval sequence. The method can be applied to cardiovascular health monitoring, intelligent health management, real-time analysis of wearable equipment and the like.

The present application discloses a natural language processing method and system based on a spiking neural network, which relates to the technical field of natural language processing. The method is based on a spiking neural network and a Transformer structure, and constructs a natural language processing model including an embedding layer, a multi-step pulse coding layer, several pulse Transformer layers and a prediction head connected in sequence, so as to fully tap the energy efficiency potential of pulse computing; then, based on the principle of maintaining a stable neuron pulse firing rate, the parameters in the natural language processing model are initialized; finally, based on a knowledge distillation algorithm, the knowledge in the artificial neural network model trained on the task set is distilled into the natural language processing model, making full use of the pre-training of the existing artificial neural network model, eliminating the need for a large-scale pre-training process from scratch, and effectively distilling the knowledge of the artificial neural network model into the spiking neural network model, thus avoiding the consumption of massive computing resources.

An abnormality detecting apparatus includes a signal obtaining unit, a first index calculation unit, a spectrum calculation unit, a standardization unit, and a second index calculation unit. The signal obtaining unit obtain signals indicating a rotation speed of the wheel assembly, as pulses having a rise. The first index calculation unit calculates a first index indicating a temporal variation of the rise of each of the pulses. The spectrum calculation unit calculates a frequency spectrum of rotational orders from a first order to an mth order of the first index calculated for each of the pulses. The standardization unit standardizes a gain for each rotational order of the frequency spectrum using a mean value and a standard deviation of the gain for each rotational order when there is no abnormality. The second index calculation unit calculates a second index for determining whether there is an abnormality.

This invention belongs to the field of machine state prediction technology, specifically relating to a fault diagnosis method and system based on an encoder-decoder attention mechanism-enhanced spiking neural network. The method includes: acquiring equipment vibration signals and segmenting them into time-series segments; using S-transform to convert each segment into a two-dimensional time-frequency image to construct a dataset; inputting the images into a lightweight hybrid network model to extract deep features. This model consists of alternating cascaded spiking convolutional encoders and spiking efficient additive attention mechanism encoders. The former extracts local features through spiking neurons and depthwise separable convolutions, while the latter uses a low-linear-complexity, efficient additive attention mechanism to model global context dependencies; finally, fault classification is completed based on the deep features. This invention, through the fusion of spiking computation and efficient attention mechanisms, significantly reduces the number of model parameters while improving feature extraction capabilities and diagnostic accuracy, making it particularly suitable for deployment on resource-constrained edge devices for fault diagnosis.

The invention provides a pulse remote signaling acquisition method and a pulse remote signaling acquisition terminal of an electric energy meter, and relates to the technical field of data acquisition. The method comprises the following steps: applying for a corresponding target storage resource for a pulse input port in a storage module; performing state scanning on the pulse input port to obtain current pulse output data of the electric energy meter; performing pulse calculation according to the pulse configuration information of the preset metering point and the current pulse output data to obtain current pulse metering data; obtaining a current remote signaling input signal of the electric energy meter collected by a remote signaling port; determining the current operation state of each controlled switch in the electric energy meter according to the current remote signaling input signal; and uploading the current pulse output data, the current pulse metering data and the current running state stored in the target storage resource to a master station system by adopting a remote communication module. Various data of the electric energy meter can be uploaded to the master station system through the pulse remote signaling acquisition terminal, and real-time monitoring of the state of the electric energy meter is realized.