102 results about "Quantization (signal processing)" patented technology

A video signal decoding device comprises a processor which: determines a particular quantizer for reconstructing a first quantized transform coefficient, the particular quantizer being one of a first quantizer and a second quantizer which are different from each other, the particular quantizer being determined on the basis of the state of the first quantized transform coefficient; reconstructs the first quantized transform coefficient on the basis of the particular quantizer to obtain a reconstructed transform coefficient; and updates the state of a second quantized transform coefficient that is reconstructed after the first quantized transform coefficient, wherein the first quantized transform coefficient and the second quantized transform coefficient are transform coefficients in the current block.

The invention discloses an audio signal processing method and device based on MOE, a medium and a product, and relates to the field of audio signal processing. According to the technical scheme, firstly, the audio signal processing device decomposes the audio signal into the sub-band signals in different frequency intervals, and the complexity parameter of each sub-band signal is calculated; on the basis of the sub-band signals and complexity parameters corresponding to the sub-band signals, the gating network in the hybrid expert model can intelligently distribute the optimal expert network combination for each sub-band signal, the speciality of each expert network is fully utilized, the processing efficiency is improved, the optimal expert network combination is screened by adopting a processing weight coefficient threshold, and the processing efficiency is improved. And redundant processing is avoided. And finally, the audio signal processing device obtains a structured audio data stream through quantization coding and data packaging. According to the dividing and treating method, the quality and efficiency of audio signal processing are improved, the processing strategy can be flexibly adjusted according to the signal characteristics, and the adaptability is high.

The invention relates to the technical field of intelligent signal processing, in particular to an unmanned aerial vehicle cluster intelligent beamforming method and system, and the method comprises the following steps: obtaining unmanned aerial vehicle node positions, antenna deviation and synchronization errors, analyzing the influence of quantization precision on beamforming, extracting low-precision waveform features, recognizing the amplitude-phase error compensation amount, and optimizing the anti-interference performance. Phase constraints are extracted and optimized, the consistency of the beam and the expected direction is compared, parameters are adjusted to obtain an adjustment instruction set, and a beam forming maintaining state label is output. According to the method, position information, pointing deviation and synchronization errors are analyzed through joint modeling, amplitude-phase feature extraction precision is improved, interference and echo separation capability is enhanced, control robustness is improved, discrete phase constraint and alternating direction optimization are introduced, control vector solution and convergence are accelerated, and direction errors are dynamically adjusted. A retention label is generated through convergence trend analysis, beam consistency and anti-interference performance are enhanced, and high-precision stable control is realized in a complex change scene.

The invention relates to the technical field of image signal processing, and discloses an agent-free ISP hyper-parameter high-speed optimization method and system, and the method comprises the following steps: collecting RAW data through a sensor, carrying out the black level correction, and initializing the ISP register parameter mapping; a multi-dimensional state vector is constructed by combining the coded image and the parameters, and the master and slave agents generate discrete adjustment amounts in parallel; after a register is updated through nonlinear quantization, a composite reward is calculated based on target detection and image quality, and data is stored and network parameters are updated through a priority experience pool; the system comprises a sensor module, an ISP hardware processing unit, a master-slave agent decision module, a reward calculation module, a register mapping module and an experience playback module. According to the method, hardware-in-the-loop and deep reinforcement learning are combined, and the problems of proxy model distortion, response lag, high data dependence and parameter mutation are solved through master-slave agent parallel decision, a self-supervision reward mechanism and nonlinear parameter mapping.

The invention relates to the technical field of vital sign anomaly detection, in particular to a vital sign anomaly feature extraction method based on a sign signal, which comprises the following steps: firstly, dynamically adjusting the size of a time cabin through variational optimization, taking both transient and steady feature capture into account, and avoiding feature loss caused by a fixed window length; in combination with principal component analysis, Lyapunov exponential chaos quantization and a high-order differential operator, nonlinear features are decoupled from multiple dimensions of a time domain, a frequency domain and a phase domain, the feature interpretability is improved, further, through phase-speed mapping and attention weighted fusion, self-adaptive enhancement and decoupling of spatial-temporal features are achieved, noise interference is suppressed, and the spatial-temporal features are enhanced. The finally generated multi-dimensional feature cube integrates time resolution, frequency precision and phase differential information, the anomaly detection robustness of complex vital signs is remarkably improved, and the defects that a traditional method is single in dimension and insufficient in adaptability in dynamic signal processing are overcome.

The invention relates to an analog-digital hybrid SRAM (Static Random Access Memory) memory audio noise reduction method and system, and the method comprises the following steps: carrying out the differential sampling quantization processing of an input audio signal, and obtaining a multi-bit-width digital audio code stream; inputting the multi-bit-width digital audio code stream into a preset SRAM (Static Random Access Memory) memory chip to carry out mixed mode write-in operation to obtain an analog-digital mixed memory matrix; performing frequency domain feature extraction on the analog-digital hybrid storage matrix to obtain frequency domain extraction features; performing noise identification and suppression on the audio signal based on the frequency domain extraction feature to obtain a time-varying noise suppression coefficient; and performing audio signal recovery processing on the time-varying noise suppression coefficient to obtain a denoised output audio signal, thereby solving the technical problem that power consumption and hardware cost are greatly increased due to the fact that a traditional digital signal processing method generally depends on a high-precision analog-to-digital converter and a high-performance processor.

The invention relates to the technical field of signal processing, and particularly provides a full-quantization spectrum characteristic statistical method, system and device and a storage medium, and the method comprises the steps: discretizing a current sampling signal frequency band into a continuous frequency point unit sequence according to a preset rule, and mapping a signal intensity value into an intensity level according to the preset rule; if the signal intensity exceeds the current noise base value, determining the frequency point range of the current sampling signal by extracting the center frequency point and the effective bandwidth of the current sampling signal, and recording the intensity level mapped for all frequency point units in the range once; if the signal intensity does not exceed the current noise base value, recording the intensity grade of mapping for one time for a frequency point unit in the frequency point unit sequence of the current sampling signal; and counting all involved frequency point units in the frequency spectrum recording process and the occurrence frequency of each intensity level recorded on each frequency point unit. According to the invention, the problem of virtual high background noise caused by a traditional peak value holding mode is solved.

Examples relate to the field of radio frequency (RF) signal processing such as classifying RF signals with low latency. The method involves receiving portions of an RF signal, transforming these portions into a time-resolved frequency representation using a continuous wavelet transform, and processing this representation with a recurrent neural network. The neural network modifies a neural network state incrementally to generate a classification output, which may include modulation classification, signal-to-noise ratio (SNR) classification, or jamming detection. The system achieves sub-millisecond inference latency through techniques such as model quantization and batch size optimization. Principal uses include real-time RF signal analysis and jamming detection, with applications in communication systems and environmental monitoring. The RF signal may be received from a quantum RF sensor based on Rydberg atoms, enabling broad frequency range detection.

The invention discloses an AI-driven vehicle-mounted sound field real-time modeling and voice separation method, and relates to the technical field of voice signal processing. A main control unit comprising a time sequence synchronizer, a resource scheduler and a health monitor is constructed. 3D sound field modeling is carried out by adopting a lightweight STCN + bidirectional LSTM network, adaptive updating of the model is realized through EWC incremental learning, and a CNN-LSTM noise classification network and targeted suppression algorithms such as ANF / spectral subtraction are developed. The voice separation module adopts an improved Conv-TasNet architecture, 3D spatial constraint and a multi-task loss function are fused, and low delay is realized under INT8 quantization and pipeline processing. The system dynamically optimizes parameters through a real-time regulation and control unit, supports scene self-adaption, finally achieves a separation effect in a mixed noise scene, reduces the delay of the whole system, and effectively improves the definition and stability of vehicle-mounted voice interaction.

The invention discloses a controllable sparse migration attack method based on generator generation, and belongs to the crossing field of communication signal processing and artificial intelligence. The method aims at solving the problems that disturbance of existing countermeasure attacks in the field of electromagnetic signals is too concentrated, concealment is insufficient, and disturbance density is difficult to control. The technical scheme comprises the following steps: simultaneously outputting disturbance amplitude and disturbance probability by using a deep generation network; classifying key regions by calculating group frequency spectrum saliency positioning pairs and generating a group mask matrix; generating a binary position matrix through a mask quantization network in combination with the group mask and the disturbance probability; training the generative network by adopting a total loss function comprising adversarial loss and smooth loss; and finally, generating a sparse controllable adversarial sample through the trained network. According to the method, the adversarial sample which is high in concealment, high in mobility and controllable in disturbance proportion can be generated, and an effective means is provided for security assessment of the deep neural network.

The application provides a pulse density modulation interface implementation method and device of an analog-to-digital converter, electronic equipment and a storage medium. The method comprises the following steps: designing a circuit architecture of an analog-to-digital converter, and setting working variable parameters for components in the circuit architecture; determining a transfer function of an output signal and a transfer function of noise based on the working variable parameters; in the case that an oversampling rate (OSR) is a first set value, a filter order is a second set value, and a quantization bit number is a third set value, calculating an optimal solution of the transfer function of the noise; determining a value corresponding to the working variable parameters of the components according to the optimal solution; and selecting corresponding components for the circuit architecture based on the value, and generating a pulse density modulation interface of an analog-to-digital converter. The application can accurately design a PDM interface, and improves the signal processing performance of the interface.

The application belongs to the technical field of radar signal processing and imaging, and provides a differential coupling single-frequency time-varying low-quantization imaging method, which comprises receiving a SAR echo signal, constructing a single-frequency time-varying threshold signal, single-bit quantization processing of a reference channel, single-bit quantization processing of a phase shift channel, double-channel data fusion and harmonic cancellation, and imaging processing based on a synthesized signal. The application effectively reduces the quantization noise base and improves the detection capability of a SAR system on weak targets under the premise of not significantly increasing the hardware cost.

The present invention belongs to the field of radio signal processing technology, and specifically relates to an infinite amplitude line spectrum estimation method based on robust compressed sensing and analog sampling ADC analog sampling when the amplitude of the signal arriving at the front end of the receiver exceeds the dynamic range of the ordinary ADC input. The analog sampling ADC analog sampling mode of the present invention and the proposed line spectrum estimation method utilize the robust compressed sensing theory to realize infinite line spectrum estimation when the amplitude of the signal arriving at the front end of the receiver is higher than the dynamic range of the ordinary ADC input, accurately estimate the amplitude of the signal and its carrier frequency information, and can overcome the problem of weak signal loss caused by direct sampling and quantization using an ordinary ADC. Experiments show that the present invention successfully estimates the amplitude and carrier frequency of multiple source signals when the input signal amplitude is higher than the ADC input dynamic range, thereby achieving reliable weak signal estimation performance at a low quantization level.

The application discloses a jitter auxiliary measurement precision enhancement method for space-based calibration, belongs to the technical field of electronic communication and signal processing, and comprises the following steps: receiving a continuous-time analog input signal from a space-based calibration system; injecting an artificial jitter signal with zero mean and subject to uniform distribution into the continuous-time analog input signal to generate a composite analog signal; performing multiple independent quantization observations on the composite analog signal by using a low-bit analog-to-digital converter, obtaining multiple quantization output values corresponding to discrete levels, calculating the arithmetic mean of the multiple quantization output values, and taking the arithmetic mean as a high-precision estimation value of the amplitude of the continuous-time analog input signal; and the jitter auxiliary measurement precision enhancement method for space-based calibration can greatly improve the measurement precision of the space-based calibration system without increasing the hardware power consumption and volume.

The invention discloses an asynchronous chaotic secret communication method and a system thereof. The transmitter obtains an estimated power threshold value according to system response of the receiver and preset calibration signal processing, and then two mode mapping sets are established; the transmitter performs self-synchronization coding on the original binary communication data to obtain coded binary communication data, and then performs encryption mapping on the coded binary communication data according to the two mode mapping sets to obtain a frequency spectrum mode sequence; the transmitter transmits a calibration signal, and the receiver sets a power threshold value according to the calibration signal; the transmitter transmits multi-mode radio frequency chaotic communication signals in different modes according to a frequency spectrum mode sequence, the receiver receives and carries out conditioning quantization to obtain recovered coded binary communication data, and then the recovered coded binary communication data are decoded to obtain recovered original binary communication data. The method is high in safety, strong in anti-interference capability, simple in system structure, low in cost, free of chaos synchronization, good in engineering realizability and capable of being used for secret communication in a complex electromagnetic environment.

This invention discloses a method for covert processing of spaceborne signals based on Tent mapping, relating to the field of signal processing. In this invention, the transmitter generates a chaotic Tent sequence, which is then transformed by Box-Muller to obtain a Gaussian-distributed first interference signal. The chaotic sequence and the first interference signal are equally divided, and an identification signal containing a synchronization flag and a checksum is constructed based on the quantization value of the first data point of each chaotic sequence segment. An amplitude adjustment coefficient is determined for each segment of the target signal, and the adjusted first interference signal of each segment is superimposed with the corresponding segment of the target signal to obtain a mixed subsequence. An identification signal is added before each mixed subsequence to form a relay signal. At the receiver, a sliding window is used to match the synchronization flag in the received signal to locate the identification signal, the quantization value at the beginning of each segment is parsed, and a second interference subsequence is independently reconstructed. A coarse estimate is obtained by performing a two-dimensional correlation between time delay and frequency offset, and the received target signal is obtained through interference subspace projection cancellation. The method proposed in this invention has low hardware overhead, strong anti-interception capability, and high cancellation accuracy.