Publication detail
Reduction of face detection false positives in mobile robot interaction using proximity sensors
KREJSA, J.
Czech title
Redukce nekorektních pozitivních detekcí obličeje pomocí proximitních snímačů během interakce mobilního robotu
English title
Reduction of face detection false positives in mobile robot interaction using proximity sensors
Type
conference paper
Language
en
Original abstract
People detection is essential for mobile robots that are suppose to communicate with human. Face detector can be used as a primary tool for such detection. The paper presents a method of combining the face detector output with distance measurements provided by low cost proximity sensors. The method reduces the false positives given by face detector, thus preventing the robot from starting the conversation when the subject is not present. Data for fusion were taken from the measurements of mobile robot Advee in real world environment. Presented approach is easy to implement on variety of platforms and is computationally cheap. Verification tests were performed with the same robot and about 30 % reduction of false positives was obtained.
Czech abstract
Detekce lidí je důležitá pro mobilní roboty které s nimi mají komunikovat. Jako primární nástroj detekce je možné použít detektor obličeje. Článek předkládá metodu, která kombinuje výstup detektoru obličeje s měřením vzdálenosti poskytovaným nízkonákladovými proximitními snímači. Metodu redukuje falešné pozitivní detekce, čímž zabraňuje tomu, aby robot spustil interakci v případě že subjekt není přítomen. Data pro fúzi byla získáná z měření mobilního robotu Advee v reálném prostředí. Prezentovaný přístup je výpočetně nenáročný a snadno implementovatelný. Verifikační testy prokázaly redukci nekorektních detekcí o cca 30%
English abstract
People detection is essential for mobile robots that are suppose to communicate with human. Face detector can be used as a primary tool for such detection. The paper presents a method of combining the face detector output with distance measurements provided by low cost proximity sensors. The method reduces the false positives given by face detector, thus preventing the robot from starting the conversation when the subject is not present. Data for fusion were taken from the measurements of mobile robot Advee in real world environment. Presented approach is easy to implement on variety of platforms and is computationally cheap. Verification tests were performed with the same robot and about 30 % reduction of false positives was obtained.
Keywords in Czech
mobilní robot, interakce člověk-robot, detekce obličeje
Keywords in English
mobile robot, human-robot interaction, face detection
RIV year
2011
Released
15.06.2011
ISBN
978-80-214-4302-0
Book
Mendel 2011
Edition number
1
Pages from–to
540–545
Pages count
6
BIBTEX
@inproceedings{BUT73090,
author="Jiří {Krejsa},
title="Reduction of face detection false positives in mobile robot interaction using proximity sensors",
booktitle="Mendel 2011",
year="2011",
month="June",
pages="540--545",
isbn="978-80-214-4302-0"
}