This paper proposes two methods for estimating number and time of arrival information of people using an impulse radio-ultra wideband radar sensor. The proposed methods utilize signal processing techniques to precisely calculate the number and range information of the people who are moving around freely in a specific area.
To estimate the number of people, the proposed people counting system uses learning-based methods. In other words, a feature vector database is constructed by exploiting the pattern of reflected signals which depend on the number of people, and then a multi-layer perceptron classifier is trained using this database. When newly received signal pulse is acquired, the system can accomplish automatic people counting through the pre-trained classifier.
To accurately acquire the range information of people, the proposed ranging system detects the course of each person. First, the system collects radar echo pulses for a time to form a frame. From the frame, the course lines of each person can be calculated using Hough transform. However, the detected course lines are contaminated due to false alarms and miss detected peaks. Therefore, we compensated the line errors by transforming each line into dot in another domain, and by clustering the dots. Finally, accurate path of each person can be estimated by conducting inverse transformation of the compensated dots to course lines.
We performed experiments in two different indoor spaces in order to validate the effectiveness of the proposed counting and ranging methods. In all experiments, the proposed methods could estimate the number and range information with higher accuracy compared to original methods. The proposed counting and ranging methods can be applied in the Internet of things services that need information about number and position of people.