46results about How to "Strong noise suppression ability" patented technology

The invention relates to a deep learning noise reduction method fusing bone vibration sensor and microphone signals. The deep learning noise reduction method comprises the following steps: enabling abone vibration sensor and a microphone to collect audio signals, and respectively obtaining a bone vibration sensor audio signal and a microphone audio signal; inputting an audio signal of the bone vibration sensor into a high-pass filtering module, and carrying out high-pass filtering; inputting the bone vibration sensor audio signal subjected to high-pass filtering or the signal subjected to frequency band widening and a microphone audio signal into a deep neural network module; and obtaining the voice after noise reduction through prediction by the deep neural network module. According to the deep learning noise reduction method fusing bone vibration sensor and microphone signals, signals of a bone vibration sensor and a traditional microphone are combined, and the strong modeling capability of the deep neural network is utilized to realize high human voice restoration degree and extremely strong noise suppression capability, and the problem of human voice extraction in a complex noise scene can be solved, and the target human voice is extracted, and the interference noise is reduced, and a single-microphone structure can be adopted to reduce the cost. In addition, the signals obtained after the frequency band of the bone vibration sensor audio signals is broadened can be directly used as output.

The invention provides an FPGA-based flux gate micro signal detecting system and an FPGA-based flux gate micro signal detecting method. The FPGA-based flux gate micro signal detecting system comprises a detecting module, an output module and a processing module which is realized by FPGA, wherein the detecting module converts a flux gate signal to a digital signal and transmits the digital signal to the processing module. The processing module performs processing for obtaining a secondary harmonic signal and transmits the secondary harmonic signal to the output module. According to the FPGA-based flux gate micro signal detecting method, after power amplification is performed on a sine wave with frequency of f0/2, saturation activation is performed on the activation coil of a flux gate sensor for generating a flux gate signal with frequency of f0. After filtering and analog-to-digital conversion are performed on the flux gate signal, cross-correlation detection is performed on the obtained digital signal and two reference signals for obtaining a secondary harmonic signal, the amplitude and the phase of the secondary harmonic signal. Finally digital-to-analog conversion is performed on the secondary harmonic signal, and the obtained analog signal drives the load to operate. According to the FPGA-based flux gate micro signal detecting system and the FPGA-based flux gate micro signal detecting method, related operation is realized by means of the FPGA. The FPGA-based flux gate micro signal detecting system and the FPGA-based flux gate micro signal detecting method have advantages of effective hysteresis reduction in signal transmission, high noise inhibition capability, convenient signal processing, and high measurement accuracy. Furthermore the FPGA-based flux gate micro signal detecting system and the FPGA-based flux gate micro signal detecting method can detect nanovolt-grade weak signals.

The invention provides an out-of-focus blurred image blind restoration method based on medium frequency. The blind restoration method includes the steps of calculating the frequency spectrum of an out-of-focus blurred image, calculating an elliptical track mean function of the frequency spectrum of the out-of-focus blurred image, carrying out variable-span smoothing filtering on the elliptical track mean function to obtain a benchmark function, calculating a medium-frequency domain of the frequency spectrum of the out-of-focus blurred image by means of the elliptical track mean function, estimating the position of a first dark ring in the medium-frequency domain by means of the elliptical track mean function and the benchmark function, calculating an out-off-focus radius through the estimated position of the first dark ring and generating an out-off-focus optical transmission function, and restoring the image through a wiener filter through the medium frequency. Compared with the prior art, the out-of-focus blurred image blind restoration method has the advantages that the out-off-radius and other parameters can be automatically identified in the out-of-focus blurred image, effective restoration of the out-of-focus blurred image is achieved, noise restraint ability is high, interaction parameters are few and restoration speed is high.

The invention discloses an intelligent plant building illumination control method and an intelligent plant building illumination control system. In the invention, a plurality of intelligent control devices and a host computer form an intelligent plant building illumination control network through an RS485 bus; each intelligent control device is arranged in an illumination control box; the intelligent control device mainly comprises a central processing unit, a lamp main circuit and a handheld remote controller and the handheld remote controller controls a plurality of intelligent control devices; the central processing unit adopted has an integrated RS-485 communication interface; and all the illumination control boxes in the plant building form the network through the RS485 bus and the network is connected to an internal Ethernet network of the plant building, so lamps can be remotely controlled and inquired through a communication program of the intelligent plant building illumination control network. The method and system can realize intelligent management of the plant building illumination, reduce waste of the energy sources, reduce labor intensity of workers and realize automatic control for the plant building illumination.

The invention relates to a channel optimization device for an electric wire high speed data communication system, which is connected between the user electric appliance and the electric wire power in series, and which includes device for filtering the phase wire and ground wire. The channel optimization device concretely includes low pass network part and filter circuit part, the low pass network part is connected with the eclectic wire power for restrain high frequency normal mode noise, the filter circuit part is connected with the user's electric appliance for restrain the common mode interference noise. Thus the invention improves the noise interference resisting ability of the electric wire high speed data communication system.

The invention discloses a self-supervision low-illumination image enhancement and denoising method based on deep learning, and belongs to the field of digital image processing. The invention aims to solve the problems that an existing self-supervised low-illumination image enhancement method based on deep learning is difficult to suppress noise and cannot directly adjust and enhance the image contrast. The method comprises a self-supervised low-illumination image enhancement network and a regular term for noise suppression, the network can be combined with any existing contrast adjustment method such as Gamma transform to realize network self-supervised training, and the noise suppression regular term can be used for a loss function during network training to enable the network to have noise suppression capability. According to the invention, the contrast and brightness of the low-illumination image can be enhanced, the color and detail information can be retained, and the noise can besignificantly suppressed. The invention can be used for enhancing and denoising the low-illumination image.

The invention discloses a reentry vehicle attitude control method based on self-adaptive gain disturbance compensation. The reentry vehicle attitude control method comprises the steps: step 1, establishing a control-oriented reentry flight kinematics and dynamics model of a reentry vehicle; step 2, establishing a fixed-time convergence disturbance compensation observer, and observing the state anddisturbance items of the reentry vehicle; and step 3, designing an integral sliding mode controller, introducing an observer disturbance observation item into the integral sliding mode controller, and performing self-adaptive adjustment on the control gain by adopting a double-layer self-adaptive gain strategy. The novel fixed-time convergence disturbance compensation observer disclosed by the invention can ensure that the observation error is quickly converged to zero within the fixed time, has good noise suppression capability, and greatly improves the anti-interference capability of an aircraft.

The invention belongs to the field of electrical impedance tomography, and particularly relates to a dynamic multi-frequency electrical impedance tomography method based on the impedance frequency spectrum constraint. The method includes the steps that 1, limited surface element dividing is carried out in an imaging region, the variety number of tissue in the imaging region and the frequency spectrum of each kind of tissue are determined, and a relation model of the volume fraction and the boundary voltage is established; 2, the boundary of the imaging region is excited through exciting currents at different frequencies, and boundary voltage data at different moments at each frequency is measured; 3, an inverse problem objective function is established according to the relation of the volume fraction change and the boundary voltage pressure obtained in the step 1 and solved, and the change value of the volume fraction in the imaging region is obtained; 4, the volume fraction change isconverted into the impedance change, the impedance change is shown in the imaging region through color levels, and an impedance change image is obtained. A novel scheme for relieving the inverse problem morbidity is provided on the aspect of increasing the observation data size, and then the space resolution and the anti-noise capacity of the reconstructed image are improved.

The invention provides a PPG signal heart rate extraction device and method based on cascaded RLS adaptive filtering. The device is composed of a PPG sensor, an acceleration sensor, a data acquisitionmodule, a data storage module, a signal analysis unit and the like. The method comprises the steps that firstly, band-pass filtering is carried out on an original PPG signal to obtain a preprocessedsignal which only contains a frequency band corresponding to the heart rate and from which baseline drifts are filtered out; then, a cascaded RLS adaptive filter is constructed to filter out motion artifacts in the PPG signal; and finally, a corresponding heart rate value is calculated by utilizing heart rate extraction and optimization algorithms. The PPG signal heart rate extraction method basedon cascaded RLS adaptive filtering can be used for filtering out the motion artifacts of the PPG signal so that the accuracy of heart rate detection can also be improved under the condition of strenuous exercise; the method can adapt to changes of PPG signal forms of different individuals and the same individual; and meanwhile, on the basis of measurement of a single-path PPG signal, wearing of auser and the popularization of the device can be facilitated.

The invention provides a total variational model non-blind restoration method and system. The method comprises the following steps: adopting a noise model to model the image degradation process; analyzing the homomorphic sub-blocks of the image, transforming the unknown parameter estimation of the noise model into the solution of linear equations; considering the sparsity of image gradient and maximum a posteriori estimation, and constructing a ROF total variational model according to the solution of unknown parameters; acquiring the luminance image, calculating the optimal solution of the ROFtotal variational model, and restoring the luminance image. This algorithm can suppress the noise and guarantee the image details and image quality at the same time. The method realizes non-blind restoration and reconstruction of biomedical images through automatic and intelligent means, and has the advantages of simple operation, high detection efficiency, high reliability and strong universality.

The invention discloses a time frequency analysis method of magnetotelluric impedance estimation. By adopting empirical mode decomposition technology of Hilbert-Huang transform, a magnetotelluric signal is decomposed to acquire mode functions of various orders for the Hilbert transform, and the time frequency spectrums of the four components of the magnetotelluric signal are acquired, and acquired instant amplitudes are used to replace Fourier spectrum of a conventional Fourier method, and then the instant amplitudes are introduced in a minimal quadratic equation to solve impedance tensors of time frequency points, and then the impedance tensor elements of the various time frequency points are calculated, and in addition, the time frequency power spectrogram of the magnetotelluric signal is converted into a time frequency impedance spectrogram. In the time frequency impedance magnetotelluric, the response function values of the various time frequency points having same instant frequencies are accumulated together, and then the acquired sum is divided by the number of the time frequency point, and the mean value thereof is a response function estimated value of a frequency value. The time frequency analysis method of the magnetotelluric impedance estimation is advantageous in that realizing of robust estimation is facilitated, and the impedance elements acquired by adopting the method conform to reality of a stratum structure, and a stronger noise inhibition capability is provided, and an estimation result is more robust.

The invention discloses an indoor temperature and humidity field rapid prediction system and method. The system mainly comprises a power supply driving module, p + 1 temperature and humidity data acquisition modules, a data transmission module and an upper computer. The temperature and humidity data acquisition module acquires temperature and humidity data. And the upper computer inputs the temperature and humidity data and the temperature and humidity predicted value into the POD mode coefficient dynamic model, calculates a POD mode coefficient, inputs the POD mode coefficient into the POD mode coefficient optimal estimator, and reconstructs an indoor temperature and humidity field. The method mainly comprises the steps that 1) the temperature and humidity data acquisition module acquirestemperature and humidity data; And 2) a dynamic model for calculating the POD mode coefficient and an optimal estimator for the POD mode coefficient are written into an upper computer. And 3) temperature and humidity data is predicted. And 4) a POD mode coefficient is calculated. 5) the POD mode coefficient is input into a POD mode coefficient optimal estimator, and establishing an indoor temperature and humidity field shown in the description; According to the indoor temperature and humidity field prediction method and system, the real-time performance is high, the precision is high, the anti-interference capability is high, and long-time prediction is achieved.

The invention relates to a target group motion feature extraction method and device based on an optimal path method, computer equipment and a computer readable storage medium, and the method comprises the steps: carrying out the one-dimensional range profile imaging of a multi-frame echo signal of a target group, and carrying out the motion compensation; estimating the position and amplitude of each scattering center by adopting a super-resolution parameter estimation method; forming a scattering center sequence by using the estimated scattering center positions, and obtaining a change track of each scattering center in the one-dimensional range profile history graph by using an optimal path method; based on function fitting, calculating separation speed characteristics of each target relative to the reference target. The method can accurately realize target group separation speed extraction, and is especially suitable for motion feature extraction of a multi-target separation scene.

The invention provides a high-order synchronous integral demodulation circuit for gyro digital signals. The high-order synchronous integral demodulation circuit comprises a signal source, a fourth-order integral circuit, a differential amplification module and a switch control module, wherein an output interface of the signal source is connected with an input interface of the four-order integral circuit through a signal source switch, and the signal source is used for outputting a signal to the four-order integral circuit; an output interface of the fourth-order integral circuit is connected with an input interface of the differential amplification module, and the fourth-order integral circuit is used for carrying out fourth-order integration on the output signal of the signal source and outputting the output signal to the differential amplification circuit; an input interface of the differential amplification module is connected with the output interface of the four-order integral circuit, and the differential amplification module is used for amplifying an output signal of the four-order integral circuit; and the switch control module controls the switch to switch the conduction path, and the switching frequency of the switch is the same as the output signal frequency of the signal source. Signal source noise can be suppressed, and signal information can be extracted.