CSI Augmentation.deformation Tutorial¶

# !pip install pysensing

In this tutorial, we will be implementing a simple csi.augmentation.deformation tutorial using the pysensing library.

import sys
sys.path.append('../..')
import pysensing.csi.augmentation.deformation as deformation
import pysensing.csi.dataset.get_dataloader as load_data
import matplotlib.pyplot as plt
import numpy as np
from scipy import signal

Load the data¶

# Define the plot function

def plot_csi_heatmap(csi_data):
    """
    Plot WiFi CSI heatmap.

    """
    data_shape = csi_data.shape

    if len(data_shape) == 2:
        csi_to_plot = csi_data
    elif len(data_shape) > 2:
        csi_to_plot = csi_data.reshape(-1, data_shape[-1])
    else:
        raise ValueError("The input data should have at least 2 dimensions.")

    plt.figure(figsize=(15, 8))
    plt.imshow(csi_to_plot, aspect='auto', cmap='viridis')
    plt.colorbar(label='CSI Amplitude')
    plt.title('WiFi CSI Heatmap')
    plt.xlabel('Time Index')
    plt.ylabel('Subcarrier Index / Flattened Antenna-Subcarrier Index')
    plt.show()


def plot_spec(csi):
    _, _, csi_spectrogram = signal.stft(csi, window='hann', nperseg=256, noverlap=128)
    specgram = np.abs(csi_spectrogram)
    plt.figure(figsize=(10,5))
    plt.imshow(specgram.squeeze(), aspect='auto', origin='lower', cmap='inferno')
    plt.colorbar(label='CSI Amplitude')
    plt.xlabel('Time (s)')
    plt.ylabel('Frequency (Hz)')

# Load the data

_, test_loader = load_data.load_classification_dataset('UT_HAR',batch_size=1)
example_csi_data_list = []
for data in test_loader:
    data, label = data
    example_csi_data_list.append(data)

example_csi_data = np.concatenate(example_csi_data_list, axis=-1)
plot_csi_heatmap(example_csi_data)

/data1/msc/zyj/yunjiao_csi/1028/yunjiao_csi/tutorials/csi_source/data/csi
using dataset: UT-HAR DATA

1. time stretching¶

Use time stretching augmentation

csi_process = deformation.time_stretching(stretch_rate=0.4)
reduced_csi_data = csi_process(example_csi_data)
plot_csi_heatmap(reduced_csi_data)

2. spectrum scaling¶

Use spectrum scaling augmentation

csi_process = deformation.spectrum_scaling(scale_factor=4)
reduced_csi_data = csi_process(example_csi_data)
plot_spec(example_csi_data.squeeze()[2])
plot_spec(reduced_csi_data.squeeze()[2])
plot_csi_heatmap(reduced_csi_data)

3. spectrum shifting¶

Use spectrum shifting augmentation

csi_process = deformation.spectrum_shifting(shift_min=0, shift_max=0.5)
reduced_csi_data = csi_process(example_csi_data)
plot_spec(example_csi_data.squeeze()[2])
plot_spec(reduced_csi_data.squeeze()[2])
plot_csi_heatmap(reduced_csi_data)