Rainfall and surface runoff effect decoupling method and device based on wireless channel state information space-time atlas and storage medium

Abstract

The invention discloses a rainfall and surface runoff effect decoupling method based on a wireless channel state information space-time atlas, and relates to the field of wireless environment perception and flood monitoring. The method comprises the following steps: acquiring multi-subcarrier channel state information (CSI) of a commercial wireless link; carrying out abnormal point elimination, amplitude calibration and phase purification on the CSI; constructing a CSI space-time atlas containing an amplitude channel and a phase channel in a preset time window according to time and subcarrier indexes; inputting the space-time atlas into a double-decoder convolution auto-encoder model, extracting mixed features by a shared encoder, reconstructing a smooth background component by a rainfall decoder, reconstructing a sparse texture component by a runoff decoder, and realizing rainfall attenuation and earth surface ponding multipath effect decoupling; rainfall intensity estimation and surface ponding detection / depth estimation are realized based on the decoupling component, and decoupled channel estimation can be used for receiving end equalization to improve communication reliability. According to the method, a prior environment geometric or physical model is not needed, and the environment sensing precision can be improved and the communication error rate can be reduced in a composite effect scene.

Description

Technical Field

[0001] This invention relates to the fields of wireless environmental sensing, flood monitoring, wireless channel state information processing, and deep learning signal processing, and particularly to a method, apparatus, and storage medium for decoupling rainfall and surface runoff effects based on a spatiotemporal map of wireless channel state information (CSI). Background Technology

[0002] Flood disasters are characterized by their suddenness, rapid evolution, and wide impact. Establishing a low-cost, high-density monitoring and early warning system that can work continuously in disaster environments is an important technical requirement for disaster reduction and prevention.

[0003] In existing technologies, flood monitoring relies heavily on specialized sensors or hydrological stations such as rain gauges and water level gauges. While these systems offer high accuracy, they also have high construction and maintenance costs, limited coverage density, and are prone to power or communication outages during extreme weather and floods, leading to data loss at critical moments.

[0004] In recent years, opportunistic environmental sensing using wireless signals in communication networks has become a new direction. Compared to using only coarse-grained metrics like RSSI, CSI can provide fine-grained information such as multi-subcarrier amplitude / phase, resulting in higher environmental sensitivity.

[0005] However, in real flood or heavy rainfall scenarios, wireless channels are often affected by both overall attenuation caused by rainfall and multipath reflection / selective fading caused by surface water accumulation. The two types of effects overlap in CSI, resulting in "same effect but different source" feature ambiguity. This makes it difficult for existing sensing methods based on single features to accurately estimate the rainfall and water accumulation status at the same time, and also makes it difficult for the communication receiver to perform effective channel compensation. Summary of the Invention

[0006] To address the aforementioned problems, the present invention aims to provide a method, apparatus, and storage medium for decoupling rainfall and surface runoff effects based on CSI spatiotemporal maps and dual-decoder autoencoders, so as to separate rainfall attenuation components and water accumulation multipath texture components, thereby improving the estimation accuracy of flood-related environmental parameters and enhancing communication reliability in harsh environments.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is: a method for decoupling rainfall and surface runoff effects based on spatiotemporal maps of wireless channel state information, including: CSI acquisition, CSI preprocessing, CSI spatiotemporal map construction, decoupling of dual-decoder convolutional autoencoder, and parameter estimation / channel equalization based on decoupling components.

[0008] Furthermore, the present invention also provides a decoupling device for rainfall and surface runoff effects, including a CSI acquisition module, a preprocessing module, a map construction module, a decoupling model module, and an output module. The device can be deployed in devices such as servers, edge computing nodes, or routers / gateways.

[0009] Furthermore, the present invention also provides an electronic device and / or a computer-readable storage medium having a computer program stored thereon for performing the above-described methods.

[0010] Compared with the prior art, the advantages of the present invention are as follows: (1) Reconstruct one-dimensional CSI time series data into a two-dimensional spatiotemporal map so that the low-frequency background changes caused by rainfall and the high-frequency sparse texture caused by water accumulation can be distinguished in structure, providing a good characterization for automated decoupling; (2) By adopting a shared encoder + dual decoder structure and combining background smoothing constraints and texture sparsity constraints, end-to-end decoupling can be achieved in the absence of pure labels, reducing the dependence on prior physical models and scene geometric information; (3) The decoupled components are used for rainfall intensity estimation and water accumulation detection / depth estimation, respectively, which can improve the perception accuracy in composite scenarios; (4) The "cleaner" channel estimation after decoupling can be used for receiver equalization, improving link robustness and reducing bit error rate in harsh environments. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the overall process and decoupling framework of the method of the present invention; Figure 2 Schematic diagram of CSI spatiotemporal spectra under different physical effects; Figure 3 This is a schematic diagram of the dual-decoder convolutional autoencoder model structure; Figure 4 This is a schematic diagram of the hardware platform and data acquisition structure for an embodiment. Figure 5 This is a schematic diagram comparing the CSI spectrum and decoupling results under the combined effect; Figure 6 This is a schematic diagram of rainfall intensity estimation based on decoupling results; Figure 7 This is a schematic diagram of the communication bit error rate (BER) versus SNR curve based on decoupled channel estimation. Detailed Implementation

[0013] The technical solution of the present invention will be further described below with reference to the accompanying drawings. The present invention is not limited to the following embodiments. Various modifications or substitutions can be made by those skilled in the art without departing from the spirit of the present invention, and all such modifications or substitutions should fall within the protection scope of the present invention.

[0014] like Figure 1 As shown, the overall process of this invention includes: data acquisition, CSI preprocessing, CSI spatiotemporal map construction, dual-decoder autoencoder decoupling, and environmental parameter estimation and communication enhancement based on decoupled components.

[0015] I. CSI Data Acquisition (Optional Example) like Figure 4 As shown, a fixed wireless link can be formed between the transmitter (TX) and receiver (RX) in the scenario, and the receiver extracts the CSI matrix from the physical layer. Simultaneously, rain gauges and water level / depth sensors can be optionally deployed to provide reference values ​​required for calibration or evaluation. This invention does not limit the specific hardware model, operating frequency band, or bandwidth, as long as it can acquire the CSI matrix containing multiple subcarriers.

[0016] II. CSI Preprocessing (1) Amplitude processing: Robust methods such as Hampel filtering are used to remove outliers from the original CSI amplitude sequence; and the amplitudes of each subcarrier in the same frame are normalized to suppress the synchronous step change introduced by AGC, so that subsequent analysis can focus more on the relative changes caused by environmental changes.

[0017] (2) Phase cleanup: The original phase is affected by the carrier frequency offset (CFO) and the sampling frequency offset (SFO), which can be modeled as: φ_{k,i}=φ_{k,i}+2π·(Δf / N)·i·τ_k+β_k+Z, Where φ_{k,i} is the true phase, Δf is the subcarrier spacing, N is the number of FFT points, τ_k and β_k represent the error terms introduced by SFO and CFO, respectively, and Z is noise. A linear regression is performed on the phase of each frame with respect to subcarrier index i to estimate the slope and intercept, which are then eliminated, resulting in a relative phase sequence that only reflects multipath variations.

[0018] (3) Subcarrier selection: Some OFDM standards have DC, protection or pilot subcarriers, and only the effective subcarriers used for data transmission can be retained to reduce interference.

[0019] III. Construction of CSI Spatiotemporal Map like Figure 2As shown, CSI frames are arranged chronologically within a time window T to construct a matrix, with the vertical axis representing time (frame order) and the horizontal axis representing the subcarrier index. Pixel values ​​can be composed of two channels, one for amplitude and one for phase.

[0020] To facilitate neural network processing, the phase can be mapped to 0~255: P_pixel=255×(phase+π) / (2π); the amplitude |H| is converted to the decibel domain A_dB=20log10|H|, and linearly normalized using the global minimum A_min and maximum A_max: A_pixel=255×(A_dB−A_min) / (A_max−A_min). This yields a multi-channel CSI spatiotemporal map of size T×N×2.

[0021] IV. Decoupling of Dual-Decoder Convolutional Autoencoder like Figure 3 As shown, the decoupled model includes a shared encoder and two parallel decoders. The shared encoder consists of stacked convolutional layers, activation layers, and pooling layers, used to extract mixed features from the input map and compress them into a latent representation z.

[0022] The rain decoder is used to reconstruct the smooth background component I_rain. Its structure can adopt transposed convolution / upsampling convolution modules, and restrict the transmission of high-frequency details by not setting skip connections, thereby biasing the learning of global, low-frequency attenuated background.

[0023] The runoff decoder is used to reconstruct the sparse texture component I_runoff. Its structure can employ transposed convolution / upsampled convolution modules and introduce skip connections to the corresponding layers of the encoder to enhance the reconstruction capability of striped high-frequency details.

[0024] To guide the separation of the two components in the absence of pure labels, a composite loss function can be used: L_total=W_rec·L_rec+W_smooth·L_smooth+ W_sparse·L_sparse.

[0025] Here, L_rec can be the mean squared error (MSE), used to constrain the reconstruction input of (I_rain + I_runoff); L_smooth can be the total variation (TV), used to constrain the gradient of I_rain to be small in the time and frequency dimensions; L_sparse can be the L1 norm, used to encourage the sparsity of I_runoff. By minimizing L_total to train the model, the decoupling of rainfall and runoff effects can be achieved in an end-to-end framework.

[0026] V. Parameter estimation and communication enhancement based on decoupled components (optional) (1) Rainfall intensity estimation: Features such as overall attenuation or average brightness can be extracted from I_rain and mapped to rainfall intensity through linear regression, support vector regression or other regression models.

[0027] (2) Water accumulation detection / depth estimation: Features such as energy (sum of squared pixels) or texture density can be extracted from I_runoff, and the existence of water accumulation can be detected by threshold discrimination. The depth of water accumulation can be output through regression model.

[0028] (3) Communication enhancement: The decoupled channel estimation corresponding to I_rain+I_runoff can be used for receiver channel estimation and equalization to suppress estimation bias caused by composite effects, thereby reducing the bit error rate.

[0029] like Figures 5 to 7 As shown, in a composite effect scenario, the present invention can effectively separate the smooth background from the sparse stripe texture and achieve synergistic improvement in environmental perception and communication performance.

[0030] It should be noted that in this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0031] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for decoupling rainfall and surface runoff effects based on spatiotemporal maps of wireless channel state information, characterized in that, Includes the following steps: Step 1: Acquire Channel State Information (CSI) data stream containing multiple subcarriers via the wireless transceiver link; Step 2: Preprocess the CSI data stream, including at least outlier removal, amplitude calibration, and phase cleanup, to obtain the cleaned CSI. Step 3: Within a preset time window, arrange the purified CSI according to time and subcarrier index to construct a two-dimensional CSI spatiotemporal map, and use amplitude and phase as different channels to form a multi-channel map; Step 4: Input the multi-channel map into the dual-decoder convolutional autoencoder model. The latent representation is obtained by the shared encoder, and the background map I_rain representing the rainfall effect is output by the rainfall decoder, and the texture map I_runoff representing the surface runoff effect is output by the runoff decoder. Step 5: Based on I_rain and I_runoff, decouple the effects of rainfall and surface runoff, and output the decoupled channel estimate I_rain+I_runoff or the corresponding environmental parameters.

2. The decoupling method as described in claim 1, characterized in that, The wireless transceiver link is a Wi-Fi, cellular communication, or other OFDM wireless link, and the CSI includes amplitude information and / or phase information of at least multiple subcarriers.

3. The decoupling method as described in claim 1, characterized in that, The outlier removal employs median-based Hampel filtering or an equivalent robust statistical method to eliminate amplitude anomalies caused by transient disturbances.

4. The decoupling method as described in claim 1, characterized in that, The amplitude calibration includes normalizing or scaling the amplitudes of all subcarriers within the same CSI frame to suppress synchronization step changes introduced by the receiver's automatic gain control (AGC).

5. The decoupling method as described in claim 1, characterized in that, The phase cleanup includes: linearly fitting the original phase of each subcarrier within each CSI frame, estimating and removing the phase offset and linear terms introduced by the carrier frequency offset (CFO) and sampling frequency offset (SFO) to obtain a relative phase sequence.

6. The decoupling method as described in claim 1, characterized in that, The construction of the CSI spatiotemporal map includes: The CSI frames continuously collected within the time window T are arranged into a matrix in chronological order, with the vertical axis corresponding to time and the horizontal axis corresponding to the effective subcarrier index. Wrap the phase value in (−π, π) and then map it to a pixel value according to P_pixel=255×(phase+π) / (2π); After converting the amplitude value to the decibel domain A_dB=20log10|H|, it is mapped to a pixel value according to A_pixel=255×(A_dB−A_min) / (A_max−A_min), where A_min and A_max are the global minimum and global maximum values ​​uniformly adopted for both the training and test sets.

7. The decoupling method as described in claim 1, characterized in that, The dual-decoder convolutional autoencoder model includes a shared encoder, a rainfall decoder, and a runoff decoder. The shared encoder consists of at least one set of convolutional layers, activation layers, and pooling layers, used to extract the mixed features of the multi-channel map and generate a latent representation; The rainfall decoder is used to reconstruct low-frequency smooth background components and does not employ skip connections from the encoder; The runoff decoder is used to reconstruct high-frequency sparse texture components and includes skip connections to the corresponding layer of the encoder to enhance detail reconstruction capabilities.

8. The decoupling method as described in claim 1, characterized in that, The training of the dual-decoder convolutional autoencoder model uses a composite loss function: L_total=W_rec·L_rec+W_smooth·L_smooth+ W_sparse·L_sparse; Where L_rec is the reconstruction loss, used to constrain the sum of I_rain and I_runoff to reconstruct the input map; L_smooth is the smoothness loss, which is applied to I_rain to penalize high-frequency changes; L_sparse is a sparsity loss applied to I_runoff to encourage sparse representation of texture components.

9. The decoupling method as described in claim 1, characterized in that, Based on the background map I_rain, the overall attenuation features are extracted and input into the regression model to output the rainfall intensity; based on the texture map I_runoff, the energy features are extracted and the existence and / or depth of surface water accumulation are output through threshold discrimination or regression model.

10. The decoupling method as described in claim 1, characterized in that, The decoupled channel estimation I_rain+I_runoff is used for receiver channel estimation and equalization processing to reduce the communication bit error rate in scenarios of combined rainfall and water accumulation.

11. A device for decoupling rainfall and surface runoff effects, characterized in that, include: The CSI acquisition module is used to acquire the CSI data stream of the wireless link; The preprocessing module is used to remove outliers, calibrate amplitude, and clean phase in CSI. The map construction module is used to construct multi-channel CSI spatiotemporal maps within a preset time window; The decoupled model module is used to output the background map I_rain and the texture map I_runoff based on the dual-decoder convolutional autoencoder. The output module is used to output the decoupled channel estimation and / or environmental parameters.

12. An electronic device and / or a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 10.

Images