CSQ_database

Full Cell Signal Quality database (test and training data) for article

S. Savazzi, S. Kianoush, V. Rampa,, U. Spagnolini "Cellular data analytics for detection and discrimination of body movements," IEEE Access, 2018

CSQ_database-master folder (CSV files)

Folder: Long_occupancy. Contains CSQ data corresponding to long occupation times (30min-2h). Check readme file in the same folder.

Folder: Short occupancy. Contains CSQ data corresponding to short occupation times (namely "entry/exit motions", 2-10min). Check readme file in the same folder.

Folder: CSQ training features. Contains the training features (hyperparameters) database. Check readme file in the same folder.

Each folder contains the following files: CID/LCID: cell indentifier corresponding to a given sample CSQ: received signal strength (dBm) corresponding to a given sample, TIME: time of sampling (HH-MM-ss) - also DATE in case of long occupation times.

True occupancy records are indicated for all cases.

CellSignals_1402-1502 folder (CSV files)

consists of a database of multicell signal quality samples, including CID information as well as accelerometer data for benchmarking. The data come from two smartphones (co-located, see the demo video https://youtu.be/8fk8zxW0g8c)   

1: xperia model (3 cells)

2: nexus (3 cells)

Folders

- 'envXXX' contains data corresponding to empty, unmodified environment 

- 'envchangeXXX' contains data corresponding to a changing environment (body and object movements, 2 targets are considered) 

- 'test_data': further data sets that can be used for performance analysis and tests

Data structure (CSV files)

Acc1, Acc2 Accelerometers data axis vs time (2400 samples, 10 samples per second), Acc1 is measured by xperia, Acc2 from nexus

CID1, CID1 Cell identifiers: time vs 3 cells

CSQ1, CSQ2, Cell signal quality: time vs 3 cells

delta_CSQ1, delta_CSQ2: CSQ deviations wrt clutter samples (obtained during app initialization, see paper): time vs 3 cells 

clutter avg CSQ for 3 cells and vs cellphones obtained during app initialization (see paper)

time: timestamps

edges: suggested edges for histogram/deviation counts computations

Body proximity discrimination folder (mat files)

Cell Signal Quality Data from two smartphones (colocated)  

smartphoneID: 1 -> xperia model (3 cells: 1 - camped cell, 2,3 - neighbor cells)

smartphoneID: 2 -> nexus (3 cells)

Files (mat files)

- 'envXXX' contains CSQ(samples, cell, smartphoneID) tensor data corresponding to empty, unmodified environment 

- 'motionXXX' contains CSQ(samples,cell,smartphoneID) tensor data corresponding to a changing environment in the surrounding of the smartphones 

- 'tamperemaXXX': contains CSQ(samples,cell,smartphoneID) tensor data corresponding to tampering attempts on smartphone 2 (body proximity to smartphone 2) 

- 'envXXX_test' contains CSQ(samples, cell, smartphoneID) tensor data corresponding to empty, unmodified environment (for validation)

- 'motionXXX_test' contains CSQ(samples,cell,smartphoneID) tensor data corresponding to a changing environment in the surrounding of the smartphones (for validation)

- 'tamperemaXXX_test': contains CSQ(samples,cell,smartphoneID) tensor data corresponding to tampering attempts on smartphone 2 (body proximity to smartphone 2) (for validation)