Help about converting ADC data to RAD data

[cuixinyuan70]

Thank you for your brilliant work first! And I am so sorry to bother you. I transform the raw ADC data with a (sample, chirps, Tx, Rx) structure to (sample, chirps, vritualRx), which is (256, 64, 8). Then do zero-padding before Azimuth FFT, making the ADC data to (256, 64, 256). Finally do 1D-FFT at the sample latitude to obtain the range, then 2D-FFT at the chirps latitude to obtain the Doppler, and 3D-FFT on the Rx latitude to obtain the angle.

The problem is that I cannot match the RA&RD heatmap generated by myself with the RAD data RADDet provided. figure_1 is generated by me, figure_2 is provided by RADDet RAD.

[cuixinyuan70]

[cuixinyuan70]

Here are my codes:

zero-padding

data_radar = np.zeros((256, 64, 256)).astype(complex) for k in range(256): for m in range(n_chirps): for n in range(8): data_radar[k, m, n] = adc_data[k, m, n]

range FFT

range_win = choose_windows('Hamming', n_samples) # add Hamming Windowing doppler_win = choose_windows('Hamming', n_chirps) range_profile = np.zeros((256, 64, 256)).astype(complex) for k in range(256): for m in range(n_chirps): temp = data_radar[:, m, k] * range_win # Hamming Windowing temp_fft = np.fft.fft(temp, N) # Do N-point FFT for each chirp range_profile[:, m, k] = temp_fft print(range_profile.shape) # (range, chirps, Rx)

doppler FFT

doppler_profile = np.zeros((256, 64, 256)).astype(complex) for k in range(256): # Rx for n in range(N): # range temp = range_profile[n, :, k] * doppler_win temp_fft = np.fft.fftshift(np.fft.fft(temp, M)) doppler_profile[n, :, k] = temp_fft print(doppler_profile.shape) # (range, doppler, Rx)

angle FFT

angle_profile = np.zeros((256, 64, 256)).astype(complex) for n in range(N): # range for m in range(M): # chirps temp = doppler_profile[n, m, :] temp_fft = np.fft.fftshift(np.fft.fft(temp, Q)) angle_profile[n, m, :] = temp_fft print(angle_profile.shape) # (range, doppler, angle)

[Acciwu]

Thank you for your brilliant work first! And I am so sorry to bother you. I transform the raw ADC data with a (sample, chirps, Tx, Rx) structure to (sample, chirps, vritualRx), which is (256, 64, 8). Then do zero-padding before Azimuth FFT, making the ADC data to (256, 64, 256). Finally do 1D-FFT at the sample latitude to obtain the range, then 2D-FFT at the chirps latitude to obtain the Doppler, and 3D-FFT on the Rx latitude to obtain the angle.

The problem is that I cannot match the RA&RD heatmap generated by myself with the RAD data RADDet provided. figure_1 is generated by me, figure_2 is provided by RADDet RAD.

Imports

import numpy as np import matplotlib.pyplot as plt

Read in frame data: ADC data

frame = np.load('./000028.npy') frame = frame.reshape((256,64,8)) frame = np.transpose(frame, (1, 2,0))

Manually cast to signed ints

frame.real = frame.real.astype(np.int16) frame.imag = frame.imag.astype(np.int16) print(f'Shape of frame: {frame.shape}')

Meta data about the data

num_chirps = 64 # Number of chirps in the frame num_samples = 256 # Number of ADC samples per chirp num_tx = 2 num_rx = 4 num_vx = num_tx * num_rx # Number of virtual antennas range_plot = np.fft.fft(frame, axis=2)

Visualize Results

plt.imshow(np.abs(range_plot.sum(1)).T) plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Range') plt.show()

---

range_doppler = np.fft.fft(range_plot, axis=0) range_doppler = np.fft.fftshift(range_doppler, axes=0)

Visualize Results

plt.imshow(np.log(np.abs(range_doppler).T).sum(1)) plt.xlabel('Doppler Bins') plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Doppler') plt.show() range_doppler.shape

num_angle_bins = 256 """ REMOVE """ padding = ((0,0), (0,num_angle_bins-range_doppler.shape[1]), (0,0)) range_azimuth = np.pad(range_doppler, padding, mode='constant') range_azimuth = np.fft.fft(range_azimuth, axis=1) range_azimuth = range_azimuth range_azimuth.shape

Visualize Results

plt.imshow(np.log(np.abs(range_azimuth).sum(0).T)) plt.gca().invert_yaxis() # Invert the y-axis plt.gca().invert_xaxis() plt.xlabel('Azimuth (Angle) Bins') plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Azimuth') plt.show()

[cuixinyuan70]

Thanks for your help! :)

[lijiunderstand]

Thank you for your brilliant work first! And I am so sorry to bother you. I transform the raw ADC data with a (sample, chirps, Tx, Rx) structure to (sample, chirps, vritualRx), which is (256, 64, 8). Then do zero-padding before Azimuth FFT, making the ADC data to (256, 64, 256). Finally do 1D-FFT at the sample latitude to obtain the range, then 2D-FFT at the chirps latitude to obtain the Doppler, and 3D-FFT on the Rx latitude to obtain the angle. The problem is that I cannot match the RA&RD heatmap generated by myself with the RAD data RADDet provided. figure_1 is generated by me, figure_2 is provided by RADDet RAD.

Imports

import numpy as np import matplotlib.pyplot as plt

Read in frame data: ADC data

frame = np.load('./000028.npy') frame = frame.reshape((256,64,8)) frame = np.transpose(frame, (1, 2,0))

Manually cast to signed ints

frame.real = frame.real.astype(np.int16) frame.imag = frame.imag.astype(np.int16) print(f'Shape of frame: {frame.shape}')

Meta data about the data

num_chirps = 64 # Number of chirps in the frame num_samples = 256 # Number of ADC samples per chirp num_tx = 2 num_rx = 4 num_vx = num_tx * num_rx # Number of virtual antennas range_plot = np.fft.fft(frame, axis=2)

Visualize Results

plt.imshow(np.abs(range_plot.sum(1)).T) plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Range') plt.show()

range_doppler = np.fft.fft(range_plot, axis=0) range_doppler = np.fft.fftshift(range_doppler, axes=0)

Visualize Results

plt.imshow(np.log(np.abs(range_doppler).T).sum(1))

plt.xlabel('Doppler Bins') plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Doppler') plt.show() range_doppler.shape

num_angle_bins = 256

""" REMOVE """ padding = ((0,0), (0,num_angle_bins-range_doppler.shape[1]), (0,0)) range_azimuth = np.pad(range_doppler, padding, mode='constant') range_azimuth = np.fft.fft(range_azimuth, axis=1) range_azimuth = range_azimuth range_azimuth.shape

Visualize Results

plt.imshow(np.log(np.abs(range_azimuth).sum(0).T)) plt.gca().invert_yaxis() # Invert the y-axis plt.gca().invert_xaxis() plt.xlabel('Azimuth (Angle) Bins') plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Azimuth') plt.show()

Hi, can you share the raw ADC data? Thank you

[lijiunderstand]

Thanks for your help! :)

Hi, can you share the ADC data? Thank you

[Acciwu]

Thank you for your brilliant work first! And I am so sorry to bother you. I transform the raw ADC data with a (sample, chirps, Tx, Rx) structure to (sample, chirps, vritualRx), which is (256, 64, 8). Then do zero-padding before Azimuth FFT, making the ADC data to (256, 64, 256). Finally do 1D-FFT at the sample latitude to obtain the range, then 2D-FFT at the chirps latitude to obtain the Doppler, and 3D-FFT on the Rx latitude to obtain the angle. The problem is that I cannot match the RA&RD heatmap generated by myself with the RAD data RADDet provided. figure_1 is generated by me, figure_2 is provided by RADDet RAD.

Imports

import numpy as np import matplotlib.pyplot as plt

Read in frame data: ADC data

frame = np.load('./000028.npy') frame = frame.reshape((256,64,8)) frame = np.transpose(frame, (1, 2,0))

Manually cast to signed ints

frame.real = frame.real.astype(np.int16) frame.imag = frame.imag.astype(np.int16) print(f'Shape of frame: {frame.shape}')

Meta data about the data

num_chirps = 64 # Number of chirps in the frame num_samples = 256 # Number of ADC samples per chirp num_tx = 2 num_rx = 4 num_vx = num_tx * num_rx # Number of virtual antennas range_plot = np.fft.fft(frame, axis=2)

Visualize Results

plt.imshow(np.abs(range_plot.sum(1)).T) plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Range') plt.show() range_doppler = np.fft.fft(range_plot, axis=0) range_doppler = np.fft.fftshift(range_doppler, axes=0)

Visualize Results

plt.imshow(np.log(np.abs(range_doppler).T).sum(1))

plt.xlabel('Doppler Bins') plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Doppler') plt.show() range_doppler.shape

num_angle_bins = 256

""" REMOVE """ padding = ((0,0), (0,num_angle_bins-range_doppler.shape[1]), (0,0)) range_azimuth = np.pad(range_doppler, padding, mode='constant') range_azimuth = np.fft.fft(range_azimuth, axis=1) range_azimuth = range_azimuth range_azimuth.shape

Visualize Results

plt.imshow(np.log(np.abs(range_azimuth).sum(0).T)) plt.gca().invert_yaxis() # Invert the y-axis plt.gca().invert_xaxis() plt.xlabel('Azimuth (Angle) Bins') plt.ylabel('Range Bins') plt.title('Interpreting a Single Frame - Azimuth') plt.show()

Hi, can you share the raw ADC data? Thank you

I'm very sorry, but after careful comparison, this program is indeed not quite right

[Acciwu]

Thanks for your help! :)

Hi, can you share the ADC data? Thank you

I only have department data, for example: https://drive.google.com/file/d/11FA6at2MzQ3g5G2T2scVBkHtLT1-9kgc/view?usp=sharing

[lijiunderstand]

Thanks for your help! :)

Hi, can you share the ADC data? Thank you

I only have department data, for example: https://drive.google.com/file/d/11FA6at2MzQ3g5G2T2scVBkHtLT1-9kgc/view?usp=sharing

Thank you very much for your help, can you share more with me as much as you can?

[Acciwu]

Thanks for your help! :)

Hi, can you share the ADC data? Thank you

I only have department data, for example: https://drive.google.com/file/d/11FA6at2MzQ3g5G2T2scVBkHtLT1-9kgc/view?usp=sharing

Thank you very much for your help, can you share more with me as much as you can?

Of course, the ADC data is stored on the Google hard drive: https://drive.google.com/file/d/1R3dalVn9_x8GIH7lm3lZIUl1ftOzoJnp/view?usp=sharing. If the RA image you processed matches the one provided by the author, can you share your code with everyone here?

[lijiunderstand]

Thanks for your help! :)

Hi, can you share the ADC data? Thank you

I only have department data, for example: https://drive.google.com/file/d/11FA6at2MzQ3g5G2T2scVBkHtLT1-9kgc/view?usp=sharing

Thank you very much for your help, can you share more with me as much as you can?

Of course, the ADC data is stored on the Google hard drive: https://drive.google.com/file/d/1R3dalVn9_x8GIH7lm3lZIUl1ftOzoJnp/view?usp=sharing. If the RA image you processed matches the one provided by the author, can you share your code with everyone here?

Sure, I will. Really appreciate your help.

[lijiunderstand]

Thanks for your help! :)

Hi, can you share the ADC data? Thank you

I only have department data, for example: https://drive.google.com/file/d/11FA6at2MzQ3g5G2T2scVBkHtLT1-9kgc/view?usp=sharing

Thank you very much for your help, can you share more with me as much as you can?

Of course, the ADC data is stored on the Google hard drive: https://drive.google.com/file/d/1R3dalVn9_x8GIH7lm3lZIUl1ftOzoJnp/view?usp=sharing. If the RA image you processed matches the one provided by the author, can you share your code with everyone here?

import numpy as np import matplotlib.pyplot as plt import math import mkl_fft import os

''' "designed_frequency": 76.8 Hz, "config_frequency": 77 Hz, "range_size": 256, "maximum_range": 50 m, "doppler_size": 64, "azimuth_size": 256, "range_resolution": 0.1953125 m/bin, "angular_resolution": 0.006135923 radian/bin, "velocity_resolution": 0.41968030701528203 (m/s)/bin '''

numSamplePerChirp = 256 numChirps = 64 range_res = 0.1953125

adc_path = "/mnt/RADDet_dataset/raw_adc/ADC/000080.npy" rad_path ="/mnt/RADDet_dataset/train/train/RAD/all/000080.npy"

read adc

frame = np.load(adc_path)

print(frame.shape) frame = frame.reshape((256,64,8))
frame = frame.transpose((0,2,1))

print(frame.shape)

raw_adc = np.zeros(frame.shape,dtype = 'complex128', order='C')

w, d,h = frame.shape

for c in range(d): temp = int((c%4)*2+c/4) raw_adc[:,c,:] = frame[:,temp,:] frame = raw_adc

mean = frame.mean(0) frame = frame - mean

range_fft = mkl_fft.fft(frame,numSamplePerChirp,axis=0)

dopplerFFTnum =64

RD_spectrums = mkl_fft.fft(range_fft,dopplerFFTnum,axis=2) RD_spectrums = np.fft.fftshift(RD_spectrums, axes=2) #(304,32,8)

RD_spectrums = np.flipud(np.fliplr(RD_spectrums)) rd = np.sum(np.abs(RD_spectrums), axis = 1)

print('rd.shape', rd.shape) plt.subplot(221)

plt.imshow(np.log10(rd)) print('RD_spectrums.shape', RD_spectrums.shape)

angle_fft_in = RD_spectrums
angleFFTnum = 256

angle_profile = mkl_fft.fft(angle_fft_in, angleFFTnum , axis =1) angle_profile = np.fft.fftshift(angle_profile, axes =1)

ra = np.sum(np.abs(angle_profile), axis = 2) print('ra.shape', ra.shape)

plt.subplot(222) plt.imshow(np.log(ra))

##################################################################

read RAD

data =np.load(rad_path) print(data.shape)

plt.subplot(223)

RD = np.sum(np.abs(data), axis=1) print('RD.shape',RD.shape) numSamplePerChirp = 256 numChirps = 64 range_res = 0.1953125 plt.imshow(np.log(RD))

RA_spectrums =np.sum(np.abs(data), axis=2) print("RA_spectrums.shape",RA_spectrums.shape) plt.subplot(224) plt.imshow(np.log(RA_spectrums)) plt.show()

Hi, here is the code for your reference.