FROM
` for i in tqdm(range(len(f[ky])), desc='Reading Records'): # reading the records

        signal = []  # ppg signal
        bp = []  # abp signal

        output_str = '10s,SBP,DBP\n'  # starting text for a new csv file

        for j in tqdm(range(len(f[f[ky][i][0]])),
                      desc='Reading Samples from Record {}/3000'.format(i + 1)):  # reading samples from records

            signal.append(f[f[ky][i][0]][j][0])  # ppg signal
            bp.append(f[f[ky][i][0]][j][1])  # abp signal`

TO

`dir_output='processed_data'
for mat_path in glob("data_uci/*.mat"):
table_name=(os.path.basename(mat_path).replace(".mat",''))
print(table_name)

with h5py.File(mat_path, 'r') as f:
    for i in tqdm(range(len(f[table_name])), desc='Reading Records'):  # reading the records
        des_ref = f[table_name][i][0]  # 3000 records
        list_hdf5record = f[des_ref][[*range(len(f[des_ref]))]]
        df_raw = pd.DataFrame(list_hdf5record, columns=['PPG', 'ABP', 'X'])
        df_raw.to_csv(os.path.join(dir_output,'%s_record_%s.csv')%(os.path.basename(mat_path),i))

`
This can boost the speed of extraction.

https://archive.ics.uci.edu/ml/machine-learning-databases/00340/data.zip

您好，请问有关于这个数据集的介绍吗？或者说我们如何获得该数据集呢

So given a new set of data where you do not know the ground truth and therefore cant have a meta file with max and min values for the ABP signal, what can you then do?

Hi, thanks for the well written code. I have an issue with normalization and denormalization, it seems that min-max normalization is used to normalize the data in data_handling.py. However, when it comes to denormalization, the algorithm only multiplies by max_value then plus min_value, shouldn't it be multiply by (max_value - min_value) + min_value? Thanks :).

Hi Nabil Ibtehaz,

I went through your paper, it's really cool.

I have a query for you that how can a raw PPG signal be given as input to your evaluate/predict code as you used already processed code by someone?

Does it work if a raw PPG signal is given as input or it needs to be preprocessed PPG as input? If it does not work with raw signals, what are the preprocessing steps need to perform on the raw signal to give as input to evaluate?

Thanks in advance,
NPK

Hi!

I downloaded the dataset from UCI repository but I found the dataset was not using wavelet decomposition( Kachuee's 2017 paper) as you mentioned in your PPG2ABP paper, actually,it was using Kachuee's 2015 preprocessing method. Do you have Kachuee's 2017 dataset?

Thanks!