HAPT Data Preparation

In this section we will see how to get the HAPT experiment data into the SFrame format expected by the activity classifier.¹

First we need to download the data from here in zip format. The code below assumes the data was unzipped into a directory named HAPT Data Set. This folder contains 3 types of files - a file containing the performed activities for each experiment, files containing the collected accelerometer samples, and files containing the collected gyroscope samples.

The first file is labels.txt, which contains activities performed for each experiment. The labels are specified by sample index ranges. For example, in experiment 1 the subject was performing activity number 5 between the 250th collected sample and the 1232th collected sample. The activities are encoded between numbers 1 and 6. We convert these to strings at the end of this section. The code below imports Turi Create, loads labels.txt into an SFrame, and defines a function to find the label given a sample index.

import turicreate as tc

data_dir = './HAPT Data Set/RawData/'

def find_label_for_containing_interval(intervals, index):
    containing_interval = intervals[:, 0][(intervals[:, 1] <= index) & (index <= intervals[:, 2])]
    if len(containing_interval) == 1:
        return containing_interval[0]

# Load labels
labels = tc.SFrame.read_csv(data_dir + 'labels.txt', delimiter=' ', header=False, verbose=False)
labels = labels.rename({'X1': 'exp_id', 'X2': 'user_id', 'X3': 'activity_id', 'X4': 'start', 'X5': 'end'})
labels

+--------+---------+-------------+-------+------+
| exp_id | user_id | activity_id | start | end  |
+--------+---------+-------------+-------+------+
|   1    |    1    |      5      |  250  | 1232 |
|   1    |    1    |      7      |  1233 | 1392 |
|   1    |    1    |      4      |  1393 | 2194 |
|   1    |    1    |      8      |  2195 | 2359 |
|   1    |    1    |      5      |  2360 | 3374 |
|   1    |    1    |      11     |  3375 | 3662 |
|   1    |    1    |      6      |  3663 | 4538 |
|   1    |    1    |      10     |  4539 | 4735 |
|   1    |    1    |      4      |  4736 | 5667 |
|   1    |    1    |      9      |  5668 | 5859 |
+--------+---------+-------------+-------+------+
[1214 rows x 5 columns]

Next, we need to get the accelerometer and gyroscope data for each experiment. For each experiment, every sensor's data is in a separate file. In the code below we load the accelerometer and gyroscope data from all experiments into a single SFrame. While loading the collected samples, we also calculate the label for each sample using our previously defined function. The final SFrame contains a column named exp_id to identify each unique sessions.

from glob import glob

acc_files = glob(data_dir + 'acc_*.txt')
gyro_files = glob(data_dir + 'gyro_*.txt')

# Load data
data = tc.SFrame()
files = zip(sorted(acc_files), sorted(gyro_files))
for acc_file, gyro_file in files:
    exp_id = int(acc_file.split('_')[1][-2:])

    # Load accel data
    sf = tc.SFrame.read_csv(acc_file, delimiter=' ', header=False, verbose=False)
    sf = sf.rename({'X1': 'acc_x', 'X2': 'acc_y', 'X3': 'acc_z'})
    sf['exp_id'] = exp_id

    # Load gyro data
    gyro_sf = tc.SFrame.read_csv(gyro_file, delimiter=' ', header=False, verbose=False)
    gyro_sf = gyro_sf.rename({'X1': 'gyro_x', 'X2': 'gyro_y', 'X3': 'gyro_z'})
    sf = sf.add_columns(gyro_sf)

    # Calc labels
    exp_labels = labels[labels['exp_id'] == exp_id][['activity_id', 'start', 'end']].to_numpy()
    sf = sf.add_row_number()
    sf['activity_id'] = sf['id'].apply(lambda x: find_label_for_containing_interval(exp_labels, x))
    sf = sf.remove_columns(['id'])

    data = data.append(sf)

Finally, we encode the labels back into a readable string format, and save the resulting SFrame.

target_map = {
    1.: 'walking',          
    2.: 'climbing_upstairs',
    3.: 'climbing_downstairs',
    4.: 'sitting',
    5.: 'standing',
    6.: 'laying'
}

# Use the same labels used in the experiment
data = data.filter_by(target_map.keys(), 'activity_id')
data['activity'] = data['activity_id'].apply(lambda x: target_map[x])
data = data.remove_column('activity_id')

data.save('hapt_data.sframe')

To learn more about the expected input format of the activity classifier please visit the advanced usage section.