Aerospace and Electronic Systems Magazine July 2018 - 48

LTE CommSense for Object Detection in Indoor Environment
this mode, the control center sets the
information and processing depending on the purpose of the LTE CommSense.
The process blocks (PBs) in the
processing chains are as follows:
1. Process-block-1 (PB-1): Channel information is extracted from
channel estimation and/or equalization blocks. Multiple UEs may
be used to increase area coverage
in the future. If multiple UEs are
used, the channel estimates of every individual UE will be passed
to PB-2.
2. Process-block-2 (PB-2): All information from individual UEs
of PB-1 are fused to pass on to
PB-3. In this article, a single piece
of UE is used. Therefore, PB-2 is
bypassed.

Figure 1.

Different stages in the processing chain of an LTE CommSense system.

3. Process-block-3 (PB-3): Information is gathered regarding the event that has to be detected.
This is passed to the ASIN block in PB-4 to train the ASIN
framework.
4. Process-block-4 (PB-4): The ASIN classifier is trained with the
information from PB-3. After that, it detects events using information from PB-1 and PB-2.
5. Process-block-5 (PB-5): The user is informed about the decision of the ASIN block. Information will mainly comprise confirmation of event occurrence and/or event details (in case the
event occurrence likelihood is estimated to be high by ASIN).
The ASIN framework [11] is used to obtain phenomenological
knowledge about the environment. ASIN is a bioinspired instrumentation scheme that aims to estimate one or a few changes only,
instead of trying to estimate everything in the scene.
First, we prove the hypothesis in simulation, and then we prove
it using practical data. To capture practical data, i.e., live LTE DL
signals from LTE base stations, the LTE receiver is implemented
in the Universal Software Radio Peripheral (USRP) N200 SDR
platform. The simulation and experiment details, data-capturing
methods, processing, and analysis are explained next.

EXPERIMENTS AND ANALYSIS
VERIFICATION IN THE SIMULATION ENVIRONMENT
To validate our hypothesis using simulation, we considered additive white Gaussian noise (AWGN) added to different ITU-specified LTE channel models [12]. Then, the corresponding CSI was
evaluated and passed to a pattern classifier to verify the feasibility
of whether the same models can be used to distinguish the channels.
48

We first modeled the LTE resource grid and then modeled
different ITU-specified channel models for LTE and applied
them, along with AWGN, to the transmitted signal. The LTE resource grid is a multicarrier OFDM signal arranged in the form
of frames. A resource element (RE) is an element of the time-
frequency grid. It is a single modulation symbol for a single subcarrier. Combined together, REs form resource blocks (RBs). A
single RB contains 12 subcarriers. For each subcarrier, 6 or 7
OFDM symbols are included. The number of OFDM symbols
depends on the cyclic prefix (CP) mode (normal or extended).
The system bandwidth in the frequency domain is an integer
multiple of RBs.
Four ITU-specified channel models were considered here: ITU
extended pedestrian A (EPA), ITU extended vehicular A (EVA),
ITU extended typical urban (ETU), and no multipath fading condition. The multipath profile of EPA, EVA, and ETU channel models
[12] are shown in Table 1. These are the types of channel conditions typically encountered for LTE telecommunication. Using
the LTE CommSense model, we tried to distinguish these types of
channel conditions. To perform this, first we determined the CSI
from the LTE receiver for different LTE channel models and fed it
to a simple nearest-neighbor (NN) classifier. Corresponding confusion matrices are also recorded.
Furthermore, we analyzed the effect of signal-to-noise ratio
(SNR) on the classification performance by observing the corresponding root mean square error vector magnitude (RMS EVM)
at different SNR values. The classification performance was observed at low SNR values ranging from −7 to 0 dB.

Simulation Setup
The highlighted blocks in Figure 1 are implemented in this analysis. A single UE is considered. Therefore, fusion of data in PB-2 is
not considered here.

IEEE A&E SYSTEMS MAGAZINE

JULY 2018

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine July 2018