This repository introduces a novel dataset for the classification of Chronic Obstructive Pulmonary Disease (COPD) patients and Healthy Controls. The Exasens dataset includes demographic information on 4 groups of saliva samples (COPD-HC-Asthma-Infected) collected in the frame of a joint research project, Exasens (, at the Research Center Borstel, BioMaterialBank Nord (Borstel, Germany).



Definition of 4 sample groups included within the Exasens dataset:

(I) Outpatients and hospitalized patients with COPD without acute respiratory infection (COPD).

(II) Outpatients and hospitalized patients with asthma without acute respiratory infections (Asthma).

(III) Patients with respiratory infections, but without COPD or asthma (Infected).

(IV) Healthy controls without COPD, asthma, or any respiratory infection (HC).

Attribute Information:

1- Diagnosis (COPD-HC-Asthma-Infected)

2- ID

3- Age

4- Gender (1=male, 0=female)

5- Smoking Status (1=Non-smoker, 2=Ex-smoker, 3=Active-smoker)

6- Saliva Permittivity:

a) Imaginary part (Min(Δ)=Absolute minimum value, Avg.(Δ)=Average)

b) Real part (Min(Δ)=Absolute minimum value, Avg.(Δ)=Average)

In case of using the introduced Exasens dataset or the proposed classification methods, please cite either of following references:

  • P. S. Zarrin, N. Roeckendorf and C. Wenger., "In-vitro Classification of Saliva Samples of COPD Patients and Healthy Controls Using Machine Learning Tools," in IEEE Access, doi: 10.1109/ACCESS.2020.3023971.

  • P.S. Zarrin, Zahari, F., Mahadevaiah, M.K. et al. Neuromorphic on-chip recognition of saliva samples of COPD and healthy controls using memristive devices. Sci Rep 10, 19742 (2020).

  • Soltani Zarrin, P.; Ibne Jamal, F.; Roeckendorf, N.; Wenger, C. Development of a Portable Dielectric Biosensor for Rapid Detection of Viscosity Variations and Its In Vitro Evaluations Using Saliva Samples of COPD Patients and Healthy Control. Healthcare 2019, 7, 11.

  • Soltani Zarrin, P.; Jamal, F.I.; Guha, S.; Wessel, J.; Kissinger, D.; Wenger, C. Design and Fabrication of a BiCMOS Dielectric Sensor for Viscosity Measurements: A Possible Solution for Early Detection of COPD. Biosensors 2018, 8, 78.

  • P.S. Zarrin and C. Wenger. Pattern Recognition for COPD Diagnostics Using an Artificial Neural Network and Its Potential Integration on Hardware-based Neuromorphic Platforms. Springer Lecture Notes in Computer Science (LNCS), Vol. 11731, pp. 284-288, 2019.


Imagine you just moved to your brand-new home and hired your energy provider. They tell you that based on the provided information they will set up a direct debit of €50/month. However, at the end of the year, that prediction was not quite accurate, and you end up paying a settlement amount of €300, or if you are lucky, they give you back some money. Either way, you will probably be disappointed with your energy provider and might consider moving on to another one. Predicting energy consumption is currently a key challenge for the energy industry as a whole.

Last Updated On: 
Tue, 07/20/2021 - 06:35

The data set is collected from MyNeuroHealth Application developed for the detection of Seizures and Falls. Data is gathered using tri-axial accelerometer placed at the upper left arm of an individual in an unconstraint environment.


The  database contains the raw range-azimuth measurements obtained from mmWave MIMO radars (IWR1843BOOST deployed in different positions around a robotic manipulator.


The database that contains the raw range-azimuth measurements obtained from mmWave MIMO radars inside a Human-Robot (HR) workspace environment. 


The database contains 5 data structures:

i) mmwave_data_test has dimension 900 x 256 x 63. Contains 900 FFT range-azimuth measurements of size 256 x 63: 256-point range samples corresponding to a max range of 11m (min range of 0.5m) and 63 angle bins, corresponding to DOA ranging from -75 to +75 degree. These data are used for testing (validation database). The corresponding labels are in label_test. Each label (from 0 to 5) corresponds to one of the 6 positions (from 1 to 6) of the operator as detailed in the image attached.


ii) mmwave_data_train has dimension 900 x 256 x 63. Contains 900 FFT range-azimuth measurements used for training. The corresponding labels are in label_train.


iii) LABELS AND CLASSES: label_test with dimension 900 x 1, contains the true labels for test data (mmwave_data_test). These are the (true) classes corresponding to integer labels from 0 to 5. Each class corresponds to a subject position in the surrounding of the robot, in particular:

CLASS (or LABEL) 0 identifies the human operator as working close-by the robot, at distance between 0.5 and 0.7 m and azimtuh 40-60 deg (positive).

CLASS 1 identifies the human operator as working close-by the robot, at distance between 0.3 and 0.5 m and azimtuh in the range -10 + 10 deg.

CLASS 2 identifies the human operator as working close-by the robot, at distance between 0.5 and 0.7 m and azimtuh 40-60 deg (negative).

CLASS 3 identifies the human operator as working at distance between 1 and 1.2 m from the robot and azimtuh 20-40 deg (negative).

CLASS 4 identifies the human operator as working close-by the robot, at distance between 0.9 and 1.1 m and azimtuh in the range -10 + 10 deg.

CLASS 5 identifies the human operator as working at distance between 1 and 1.2 m from the robot and azimtuh 20-40 deg (positive).


iv) label_train with dimension 900 x 1, contains the true labels for train data (mmwave_data_train), namely classes (true labels) correspond to integers from 0 to 5. 


v) p (1 x 900) contains the chosen random permutation for data partition among nodes/device and federated learnig simulation (see python code).


This dataset consists of RSS data measured from smartphones carried by two human beings.


Two users were standing at a certain distance (d = {0.2:0.2:2, 3:5}) from each other. For each distance, the App was made to scan for incoming BLE signals for about 1min. The following information is logged: the truth distance, name of smartphone, MAC address of BLE chipset, the packet payload, RSS values, time elapsed, and timestamp.  Two phones were used: 1) gryphonelab, and 2) HTC One M9.


We consolidated the data and reorganized them while applying moving average to the raw RSS value.  The reorganized data has the following format:

  • ·         device name,
  • ·         time elapsed,
  • ·         rss (raw RSS value),
  • ·         mRSS10 (filtered RSS value with window size = 10),
  • ·         mRSS100 (filtered RSS value with window size = 100),
  • ·         distance, and
  • ·         label

This paper presents a novel implementation scheme

of the essential circuit blocks for high performance, full-precision

Booth multipliers leveraging a hybrid logic style. By exploiting

the behavior of parasitic capacitance of MOSFETs, a carefully

engineered design style is employed to reduce dynamic power dissipation

while improving the glitch immunity of the circuit blocks.

The circuit-level techniques along with the proposed signal-flow

optimization scheme prevent the generation and propagation


This dataset provides the magneto-inertial signals from six MIMU (2 Xsens, 2 APDM, 2 Shimmer) and orientation from 8 reflective markers (VICON) at 3 different speeds (slow, medium, fast). Marker trajectories are provided. Proprietary orientations from MIMU vendors are also included. All data are synchronized at 100 Hz.


Including the symbol under demodulation in data-aided reference/pilot recovery process, dominates the contribution of all the received sybmols. Therefore, excluding the contribution of the symbol under detection allows other symbols to equally contributeto the reference estimation.


Each folder is named by its corresponding figure.


The Here East Digital Twin was a six month trial of a real-time 3D data visualisation platform, designed for the purpose of supporting operational management in the built environment. 


1. View the project video for context:

2. Review details of the Envirosensor implementation:

3. One option for analysing the data is using Spark and Python in a Jupyter Notebook as outlines in the following example GitHub repository:


Truth discovery techniques, which can obtain accurate aggregation results based on the weighted sensory data of users, are widely adopted in industrial sensing systems. However, there are some privacy matters that cannot be ignored in truth discovery process. While most of the existing privacy preserving truth discovery methods focus on the privacy of sensory data, they may neglect to protect the privacy of another equally important information, the tagged location information.