DIPteam has successfully completed a list of projects in the field of digital video and audio processing and analysis.
View some of the video and audio projects below.
Duration: 2020 - present
Funded by EBRD and Public Enterprise for State Roads, Skopje, North Macedonia.
The project considers the design and implementation of a Road Traffic Analysis System that is based on the video surveillance system of the Miladinovci - Stip motorway section. The system will enable traffic participants and events detection and recognition. Real time traffic status and events detection will enable fast response and resolution of potentially dangerous situations. The information collected by the system will enable efficient Intelligent Transportation Systems design and implementation, and road utilization and maintenance analysis.
View a demo of the traffic analysis and monitoring software here ....
Duration: 2020 - present
Funded by EBRD and Public Enterprise for State Roads, Skopje, North Macedonia.
The project considers the design and implementation of a Road Tax Payment Control System that utilizes an automated plate recognition system for vehicle detection and localization. The plate recognition system is based on video streams from the video surveillance systems installed in the Toll collection stations in the country and other video surveillance systems operated by PESR.
View some results of the licence plate recognition here ....
Duration: 2020-2021
Funded by Alfa-Zet Systems, Heusden-Zolder, Belgium.
The project considers the design of image analysis software capable of real-time product recognition. It is intended to be used in unmanned self-service restaurants of tomorrow, where the counter is nonexistent and the process of product recognition and counting is fully automated at the check out booth and the customer's account is automatically charged.
View some results of the product detection algorithm here ....
Duration: 2019 - 2021
Funded by iTek Systems, Gevgelija, and supported by Fund for innovations and technology development, North Macedonia.
The goal of the project is the design and implementation of a prototype of the traffic objects and events detection system that is used for traffic monitoring and control. The system utilizes information from an array of cameras and microphones mounted in specific road locations. The computer system performs information analysis in order to detect and localize traffic participants and to estimate their motion parameters. The traffic events detection system detects and recognizes specific events. The set of events includes overspeed violation, lane change violations, overtaking violations, parking violations, wrong direction driving, traffic accidents, presence of pedestrians and objects in the traffic lanes, and traffic congestion in crossroads and critical road segments.
View the research paper on traffic analysis here...
View the research paper on traffic sound detection here...
View a demo of the traffic analysis and monitoring software here ....
View some results of sound-based car crash detection here ....
Duration: 2017 - 2019
Funded by Fast Global Solutions, Inc., Glenwood, MN, USA.
The project considers the design of a 3D scene analysis algorithm for object recognition and localization based on LIDAR and image data. The algorithm is intended to be used for automated trailer loading/unloading. The goal is recognition and localization of rectangular shape packages, transport bags and objects of irregular shape. The outcome of the algorithm is dimensions, absolute position and orientation of the objects in the 3D scene, in order to enable robots to decide and perform successive object removal without causing tumbling down and damaging of the unmoved objects. The computational complexity of the algorithm should enable its real-time application.
View the research paper on package recognition here...
View the research paper on shipping bag segmentation here...
View some results of package recognition here ....
Duration: 2017 - 2020
Funded by Mikrosam A.D., Prilep, North Macedonia.
Project goals: real time quality control of automated fiber placement based on profile and image data. The algorithm utilizes laser profile scans and image data in order to detect defects during the fiber placement in real time and enable stopping of the process and correction of detected errors.
Duration: 01.01.2012 - 31.12.2012
Funded by Texas Instruments Incorporated, Dallas, Texas.
Duration: 01.01.2011 - 30.06.2012
Funded by Faculty of Electrical Engineering and Information Technologies, Ss. Cyril and Methodius University, Skopje, Macedonia.
Although the problem of motion estimation errors is thoroughly studied in the last years, the offered solutions give acceptable results only in the cases of severe motion estimation errors, and when the input images are of the specific types that can be accurately modelled. In a general, real-world case, these solutions are ineffective and result in SR-image with relatively high blur and visual artifacts. To overcome this problem, a new approach is necessary, which will result in new methods and techniques.
The research in the framework of this project should offer solutions for the mentioned problems. The research on reduction of the influence of the motion estimation errors will produce SR-procedures that are robust on these errors. The team for the proposed project performed a preliminary survey, which shows that there is a possibility to create motion estimation techniques adapted to the requirements of the SR-procedures.
Duration: 01.11.2007 - 30.04.2010
Research project in cooperation with NXP Software, B.V., Eindhoven, Netherlands.
The aim of this research project is to investigate algorithms and techniques in the following areas:
This research program was focused on new and innovative algorithms for the above mentioned fields that will take into account the resources and performances of mobile platforms.
Duration: 2020-2021
Funded by Retell Limited, Sunbury on Thames, Middlesex, England.
Design and implementation of speaker identification algorithm based on audio sample of recorded telephone conversation. The algorithm enables accurate and reliable identification of the speaker among a given set of pre-recorded speakers with time delay small enough for near real time implementation.
Duration: 01.04.2014 - 31.03.2016
Funded by Swiss National Science Foundation (SNSF) and the Swiss Agency for Development and Cooperation (SDC).
The DIPteam has joined forces to do top-notch research on Speech Prosody with three other partners: Idiap Research Institute, Switzerland, the Budapest University of Technology and Economics, Hungary, and the University of Novi Sad, Serbia. The focus of the project is the development of a unified prosody model that will be trans-lingual on the pan-European level.
Duration: 21.12.2011 - 12.12.2014
Research project on spectro-temporal features for noise robust Automatic Speech Recognition
This is a part of the research Branislav Gerazov did in his PhD Thesis. The Kernel Power flow Orientation Coefficients (KPOCs) are a novel feature set based on spectro-temporal analysis that uses a bank of 2D kernels to extract the dominant orientation of the power flow at each point in the auditory spectrogram of the speech signal. The collection of dominant power flow orientation angles forms a novel representation of the speech signal named the Power flow Orientation Spectrogram (POS), which is innately resistant to the spectral masking introduced by the presence of noise and reverberation. This approach not only grants KPOC its noise robustness, but also keeps the number of output coefficients inherently small, thus eliminating the need of the feature dimensionality reduction otherwise necessary in the conventional the spectro-temporal approach. KPOCs performance has been evaluated on three experimental frameworks, and the results have shown that they outperform a number of well-known noise-robust features for average and low SNRs.