Non-Line-of-Sight Vehicle Localization based on Sound

T Junction with a wall, Left particles and sound paths

T Junction with a wall, Right particles and sound paths

T Junction without a wall, Left particles and sound paths

T Junction without a wall, Right particles and sound paths

---

Qualitative result

Quantitative results in ARIL dataset

ARIL	Left	Right
T Junction with wall
T Junction without wall

Quantitative results in OVAD dataset

OVAD	Left	Right
T Junction with wall
T Junction without wall

---

Dataset

The dataset is available at here.

Folder Structure (Recommended)

${ROOT}
├── ARILDataset                 # Dataset
|    ├── SA                     # T-Junction with a wall
|    |  ├── left                # Direction of travel for NLoS vehicle
|    |  |  ├── [Scenario name]
|    |  |  | ├── .wav           # sound file
|    |  |  | ├── .csv           # result of SRP-PHAT algorithm
|    |  |  | └── .xlsx          # ground_truth
|    |  └── right  
|    └── SB                     # T-Junction without a wall
├── config
└── utils

---

How to run the code

SRP-PHAT algorithm running first

sh featureExtract.sh

For localization test

sh localization_test.sh

For making azimuth map

python azimuth.py

For calculating RMSE

python rmse.py

Authors

copyright
Autonomous Robot Intelligence Lab, SNU

Mingu Jeon
Jae-Kyung Cho
Hee-Yeun Kim
Byeonggyu Park
Seung-Woo Seo
Seong-Woo Kim