https://www.um.edu.mt/library/oar/handle/123456789/34630 Sun, 03 May 2026 13:29:38 GMT 2026-05-03T13:29:38Z https://www.um.edu.mt/library/oar/handle/123456789/70226 Title: Face photo-sketch recognition using deeply-learned and engineered features Abstract: Face sketches created from eyewitness descriptions of criminals have proven to be useful in assisting law enforcement agencies to apprehend perpetrators, particularly in cases lacking evidence. These sketches are typically disseminated to the public and law enforcement officers so that any persons recognising the suspect in the sketch may come forward with information leading to an arrest. However, this process is time consuming and not guaranteed to be successful. In this dissertation, an investigation of popular and state-of-the-art face photo-sketch synthesis and recognition methods which can identify perpetrators automatically is performed using an evaluation set-up that reflects real-world scenarios, through the use of challenging sketches and an extended gallery which simulates the extensive mugshot galleries maintained by law enforcement agencies. The University of Malta Software-Generated Face Sketch (UoM-SGFS) database was also created to enable the design and evaluation of algorithms when using software-generated sketches, that are nowadays being used more often than hand-drawn sketches. This database is the largest software-generated face sketch database, one of the few containing multiple sketches per subject, and the only one containing sketches represented in colour. Several novel methods have also been designed and evaluated, namely: (i) the Eigenpatches (EP) approach which improves upon the performance of the popular Eigentransformation (ET) method by transforming photos into sketches or sketches into photos on a local level, (ii) the log-Gabor-MLBP-SROCC (LGMS) method that extracts modality-invariant features, (iii) the DEEP (face) Photo- Sketch System (DEEPS) framework that applies transfer learning to a state-of-the- art face recognition system based on a Deep Convolutional Neural Network (DCNN) with the aid of an extensive set of synthetic images created using a 3D morphable model, (iv) the use of multiple synthetic sketches during system deployment, and (v) the fusion of intra- and inter-modality methods which are shown to be capable of providing complementary information. The finalised system fuses LGMS with DEEPS to yield a system outperforming state-of-the-art methods for all types of sketches, including real-world forensic sketches. Moreover, the proposed approach is efficient in terms of both computation time and template size, thereby permitting its implementation in the real-world. Description: PH.D. Mon, 01 Jan 2018 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/70226 2018-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/70164 Title: Cross-layer design for multi-view video plus depth transmission over LTE networks in crowd event scenarios Abstract: The development of digital multimedia systems has seen an unprecedented growth in recent years, with immersive video technology taking a central role. An application which is driving interest in this technology is free-viewpoint video which allows viewers to interactively navigate a scene by selecting their preferred viewing position. This is made possible through the generation of novel viewpoints rendered from a small set of texture and depth map views using a view synthesis technique. Meanwhile, according to Cisco, mobile video traffic accounted for 60% of total mobile data traffic in 2016 and this is expected to reach 78% by 2021. This growth in mobile video traffic coupled with the introduction of free-viewpoint video in the mobile ecosystem will have an impact on the user experience especially in crowd event scenarios. Such scenarios are characterised by high uplink user data traffic coupled with excessive uplink signalling overhead caused by channel quality feedback reports. In this thesis, the high uplink signalling overhead problem is tackled through the design and development of a set of novel Channel Quality Indicator (CQI) feedback reduction schemes. These are based on a User Equipment (UE)-assisted predictive filtering technique and a CQI clustering scheme respectively, where the latter is able to achieve an uplink signalling feedback reduction of 88.2%. Moreover, a cross-layer depth-texture bit rate allocation estimation technique and an enhanced depth map rate control scheme aimed at improving the synthesised view quality is proposed. Furthermore, a content-aware scheduling algorithm based on the widely used modified largest weighted delay first (M-LWDF) packet scheduling scheme is designed and tested in conjunction with the combined CQI feedback reduction schemes mentioned above. Whilst, the content-aware scheduling scheme yields an improvement in both the system performance and visual quality metrics, the use of the feedback reduction schemes has a detrimental effect on the visual quality. For this reason, a lean cross layer technique is designed to adapt the CQI feedback by soliciting CQI reports from individual UEs. This solution has not only improved the texture and synthesised view Peak Signal-to-Noise Ratio (PSNR) quality, approaching that of the content-aware MLWDF scheme without any CQI feedback reduction applied, but also achieves an uplink feedback signalling overhead reduction of 84.1%. Description: PH.D. Mon, 01 Jan 2018 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/70164 2018-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/35423 Title: Development of an augmentative and alternative communication app for the Maltese language Abstract: Augmentative and Alternative Communication (AAC) embodies all methods of communication serving as an alternative to speech. Maltese children having complex communication needs, use various AAC devices on a daily basis. Their conversation skills are mainly limited by two key factors. Firstly, AAC users communicate up to 20 times slower than people who use speech as their primary method of communication. Secondly, an AAC app for the Maltese language is currently unavailable. The aim of this work was to overcome these two limitations through the development of an AAC app targeted for the Maltese language, which provides an intelligent word suggestion mechanism to improve AAC communication rates. The app is based on a trigram language model which is able to predict the subsequent word required by the user, by considering the two previously selected words. The model was trained by means of a corpus which was specifically created for this project and uses the Interpolated Kneser-Ney smoothing technique in order to correctly resolve contexts which were not observed during training. The app enables users to retrain and update the language model, such that it may provide additional personalised word suggestions. The app was evaluated by a number of clinicians and educators who regularly work with AAC users. They remarked that it will be potentially helpful in aiding Maltese children during intervention sessions, due to its effective features. The underlying language model features an average perplexity of 90:47 when tested with non-similar training and test data and an average perplexity of 3:61 when evaluated for highly similar training and test data. The low perplexity values suggest that the language model employed in this app is remarkably accurate, and effectively performs as other trigram language models reported in literature. Description: B.SC.(HONS)COMPUTER ENG. Mon, 01 Jan 2018 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/35423 2018-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/35422 Title: FPGA-based phase control for indoor light regulation Abstract: Automation can be described as a device controlling a process without human assistance. Home automation is the future, where a person may remain comfortable and do tasks effortlessly. Nowadays, technology has improved the efficiency of automation circuitry. In this scenario, the subject is ambient light in a room where, with appropriate control, light inside a room may be adjusted to the user’s desires. Smart Glass is used, mainly focusing on polymer dispersed liquid crystal (PDLC) film, to maintain the user’s required illumination level using sunlight, while if outdoor lighting is insufficient, a lightbulb inside the room will switch on to accommodate the user. Controlling the Smart Glass uses minimal power, averaging about 5W/m2, which may be more economically and environmentally friendly than lighting a room with light emitting diode (LED) lightbulbs. Phase control on the film and lightbulb may be performed to manage the average power delivered to these devices, using a triac. Delaying the trigger pulse to its gate will manipulate the alternate current (ac) voltage wave passing through the device, which results in less power being delivered to the load. The main controller adopted is a Field- Programmable Gate Array (FPGA), which takes care of all operations, including sensor reading and phase control. The user additionally controls the lighting inside a room through a smartphone application, while automation is used to keep the illuminance as requested by the user. Results demonstrate that the project is a success, and the light to be controlled is no more than about 2000 lux. This limitation is attributed to the smart glass technology used, but the project is proof of a concept which can be adapted to other technologies, which may be more suitable for exposure to greater illuminance. Description: B.SC.(HONS)COMPUTER ENG. Mon, 01 Jan 2018 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/35422 2018-01-01T00:00:00Z