摘要:
A gas recognition method based on a compressive sensing theory. The method comprises: collecting compressed data in an under-sampling manner; performing a reconstruction on the collected compressed data to obtain reconstructed data; training a back-propagation neural network by using the reconstructed data and storing the trained back-propagation neural network; inputting data under test into the trained back-propagation neural network, such that the trained back-propagation neural network performs a recognition on the data under test to realize qualitative recognition of gas. The method solves the problem in transmission and storage of large amount of data and the problem of imprecise recognition in current gas detection, and achieves the object that a precise qualitative recognition is achieved by using a reduced amount of data.
摘要:
A sampler adapted to a one-dimension slow-varying signal, including: a signal preprocessing unit configured to preprocess an input signal; a slope-controllable sawtooth wave signal generating unit configured to generate a slope-controllable sawtooth wave signal and perform zero-resetting; a signal comparing unit configured to compare the preprocessed input signal from the signal preprocessing unit with the sawtooth wave signal and to output a pulse signal to the generating unit and a signal outputting unit when the preprocessed input signal is equal to the sawtooth wave signal; a counting unit configured to count a number of clock signals while the sawtooth wave signal generating unit is generating the sawtooth wave signal and to transmit the counted number to the signal outputting unit; the signal outputting unit configured to, upon receipt of the pulse signal output from the signal comparing unit, output the number counted by the counting unit at the moment.
摘要:
Disclosed is a neural network operation device, including: an operation array including operation units, wherein each operation unit includes: a source terminal, a drain terminal, a gate electrode, a threshold voltage adjustment layer under the gate electrode, and a channel region extending between a source region and a drain region, the threshold voltage adjustment layer is located on the channel region. The gate electrodes of each column of operation units of the operation array are connected together, and each column is used to adjust a weight value according to a threshold voltage adjusted by the threshold voltage adjustment layer. The threshold voltage adjustment layer is a ferroelectric layer.
摘要:
A writing method and erasing method of a fusion memory are provided, and the fusion memory includes a plurality of memory cells, and each memory cell of the plurality of memory cells includes a bulk substrate; a source and a drain on the bulk substrate, a channel region extending between the source and the drain, and a ferroelectric layer and a gate stacked on the channel region; and the writing method includes: applying a first voltage between the gate of at least one memory cell and the bulk of at least one memory cell, in which the first voltage is less than a reversal voltage at which the ferroelectric layer is polarization reversed, and each of the source and the drain is grounded or in a floating state.
摘要:
A transition metal oxide based selector, a method for preparing the same and resistive random access memory are provided. The method comprises: S1, forming a tungsten plug on a transistor; S2, using the tungsten plug to function as a lower electrode, and preparing a transition metal layer on the tungsten plug; S3, oxidizing the transition metal layer to convert the transition metal layer into a transition metal oxide layer; and S4, depositing an upper electrode on the transition metal oxide, patterning the upper electrode and the transition metal oxide. The selector of the present disclosure may provide a high current density and has a good uniformity. The formed 1S1R structure may effectively eliminate crosstalk phenomenon in a resistive random access memory array, and effectively increase the storage density without increasing the storage unit area, thereby increasing device integration. In addition, the selector for the resistive random access memory of the present invention has advantages of a simple structure, easy for integration, a low cost, a good uniformity, and compatibility with a CMOS process.
摘要:
A method for collecting a signal with a frequency lower than a Nyquist frequency includes, by a data transmitting end, selecting a suitable transformation base matrix for an input signal, deriving a sparse representation of the signal using the transformation base matrix to determine a sparsity of the signal, calculating a number M of compressive sampling operations according to the sparsity, sampling the signal with fNYQ/M using M channels, and integrating sampling values of each channel to obtain M measurement values. A reconstruction end reconstructs an original signal by solving optimization problems. Based on theory, compressive sampling can be performed on a sparse signal or a signal represented in a sparse manner with a frequency much lower than the Nyquist frequency, overcoming restrictions of the typical Nyquist sampling theorem. The method can be implemented simply and decrease pressure on data collection, storage, transmission and processing.
摘要:
Disclosed is a memory, including a plurality of memory units, wherein each memory unit includes: a bulk substrate; a source electrode, a drain electrode and a channel region extending between a source region and a drain region that are located on the bulk substrate; a deep-level defect dielectric layer on the channel region; and a gate electrode on the deep-level defect dielectric layer. The memory of the present disclosure allows the memory unit to operate in the charge trapping mode and the polarization inversion mode. Therefore, the memory has functions of both DRAM and NAND, and combines the advantages of the two.
摘要:
Provided are a symmetric memory cell and a BNN circuit. The symmetric memory cell includes a first complementary structure and a second complementary structure, the second complementary structure being symmetrically connected to the first complementary structure in a first direction, wherein the first complementary structure includes a first control transistor configured to be connected to the second complementary structure, the second complementary structure includes a second control transistor, a drain electrode of the second control transistor and a drain electrode of the first control transistor being symmetrically arranged in the first direction and connected to a bit line, and the symmetric memory cell is configured to store a weight value 1 or 0.
摘要:
The present disclosure provides a self-rectifying resistive memory, including: a lower electrode; a resistive material layer formed on the lower electrode and used as a storage medium; a barrier layer formed on the resistive material layer and using a semiconductor material or an insulating material; and an upper electrode formed on the barrier layer to achieve Schottky contact with the material of the barrier layer; wherein, the Schottky contact between the upper electrode and the material of the barrier layer is used to realize self-rectification of the self-rectifying resistive memory. Thus, no additional gate transistor or diode is required as the gate unit. In addition, because the device has self-rectifying characteristics, it is capable of suppressing read crosstalk in the cross-array.
摘要:
A selector for a bipolar resistive random access memory and a method for fabricating the selector are provided. The method includes: providing a substrate; forming a lower electrode on the substrate, where the lower electrode is made of a metal, and the metal is made up of metal atoms which diffuse under an annealing condition of below 400° C.; forming a first metal oxide layer on the lower electrode; performing an annealing process on the first metal oxide layer to make the metal atoms in the lower electrode diffuse into the first metal oxide layer to form a first metal oxide layer doped with metal atoms; forming a second metal oxide layer on the first metal oxide layer doped with metal atoms; forming an upper electrode layer on the second metal oxide layer; and patterning the upper electrode layer to form an upper electrode.