Abstract
      
     
      Blind and impaired people need for an assistance device able to provide them independent mobility. The design and development of such a navigation device will mean a significant advance in engineering and research. In these previous decades, many researchers have been investigating on different methods for environment information representation, able to be implemented in electronic travel aids for blind and impaired people.
      The present thesis carries out the design, modelling, implementation, experimentation and analysis of a wearable object detector and navigation device for blind people, named Cognitive Aid System for Blind People, CASBliP. The CASBliP device represents an Electronic Travel Aid, whose primary goal is to help blind users to navigate independently and safety both in indoor and outdoor environments. 
      In this context, the thesis begins with a detailed presentation of the state-of-the-art of the nowadays existing electronic travel aid systems for blind people, as well as those under development. This review comprises devices developed from the Second World War, when the development of sensors played an important role in the human life, until nowadays. In this initial chapter, a classification of the electronic travel aid systems into three main groups, based on the type of system, is presented: obstacle detectors, environmental sensors and navigation systems. More than forty relevant systems are described, explaining the main differences between the different categories. Despite the increasing knowledge and wide usage of the sensory electronic travel aid systems in the world, the development of a universal and more accurate navigation and object detection system has not been achieved yet. 
      In order to achieve the thesis objective, the designed device consists of two inputs, one output, a portable computer and a FPGA as processing units, which can function individually. The input system consist of an array of 64?1 CMOS Time of Flight sensors attached to a pair of glasses and two cameras mounted on a helmet. The system output consists of a pair of mini stereo headphones through which the user will perceive the environmental objects and free paths. The input system goal is to capture environmental information from the user’s direction of view. Taking as a basis the 3D environmental information perceived, the input system represents the moving objects and detects all static and moving objects and free paths using the depth maps, segmentations algorithms and motion detection algorithms. The high resolution input of the image is projected onto high resolution acoustic sounds via a methodology based on navigation criteria and models based on convolution with non-individual Head-Related Transfer Functions.
      The device implements a methodology for simulating that a series of sounds, in a virtual way, are radiated by the user surrounding objects, sounds which are capable to carry accurate spatial information. The idea is to generate in the user a correct perception of virtual sound sources emitting from the object surfaces, which aims to allow the brain to create a three-dimensional perceptual image of those objects like in the real world. Using these ideas, it intends to create a global perception of the sound, enabling blind people, on real-time, to perceive and get a global image of the surrounding environment and the way the objects are organized. 
      It is well known that humans use a wide range of information for navigation as vision, feelings and hearing. When human visual system is damaged (loss of vision), the hearing system gets the main role on navigation. It is extremely important to analyze and define the aspects of the visual scene, which represent the most important features for navigation and object identification, in order to represent the object presence and determine its position in space. 
      In the third chapter, an overview of the auditory system is described including its basic components and auditory organisation. This chapter provides a background of sound localization with acoustic cues (monaural and binaural cues, interaural time difference and interaural level difference, reverberation effect, cone of confusion, precedence effect and cross-correlation model). This review is aimed to represent the whole picture of the level of accuracy for sound localization. This chapter introduces the bases of the next chapter, where the properties of the sound source localization are analyzed.
      In order to achieve the goal of the thesis -the necessity of generating acoustic maps for the detected object representation and once the foundations of the auditory system and auditory factors which contribute to the sound source localization have been presented, two methods for the human spatial hearing and sound localization in situations involving multiple sound sources are explained in the Chapter four. The developed method is based on the application of the non-individual Head-Related Transfer Function to static and moving sound source localization through headphones. Against other methods based on the sound source localization using the non-individual Head-Related Transfer Functions, the approach developed in this thesis is based on the study of the evolution of the time delay between two characteristic sounds and its importance in the sound source localization through headphones. Unlike other methods, which analyse the sound source localization and sound parameters directly in the anechoic chamber, where the users localize the sounds delivered by the system through speakerphones or headphones, the proposed approach analyses the sound source localization and its parameters in off-line. The Head-Related Transfer Functions are calculated and measured using a KEMAR manikin and later convolved with a sound through a computer program. 
      Two sets of experiments are described according to the examined spatial performance involving simple broad-band stimuli. Both experiments measured how well single and train of static and moving sounds are localized in laboratory conditions, for future implementation in the navigation system. These experiments demonstrated that sound source is essential for accurate three-dimensional localization. The approach was based on presenting the sounds as overlapped in time, in order to observe the performance in localization; the objective was to see how time delay between two sounds (inter-click interval) influences on sound source localization. It was found that better localization performance was achieved for trains of sounds. Moreover, the sound perception threshold was studied. In the second study the localization of a moving sound source both in distance and azimuth was analyzed. The results demonstrate that the best results were achieved for an inter-click interval of 150ms. When comparing the localization accuracy in distance and azimuth, better results were obtained in azimuth. Also, it was noted that spatial cues such as interaural time difference and interaural level difference play an important role in spatial localization. The interaural cues arise due to the separation of the two ears, and provide information about the lateral position of the sound. 
      In the fifth chapter a series of experiments was conducted on blind people, in order to measure their performance in object detection and localization involving one or multiple sound sources. The general approach was to present various objects in order to observe the performance in object detection and sound externalization in various situations. Furthermore, the navigation and object detection and localization was tested in different scenarios. The object localization through acoustical signals arises due to the understanding of listened sounds externalization, which provides information about the spatial position of the source. 
      Three sets of experiments and two preliminary tests are developed; they make use of the acoustical object detection and navigation system. The two preliminary tests carried out the performance of the sound and object localization accuracy, with the mean to see how the end-user manage the system, how he perceive the system functionality, to analyze which components of the system must be improved or changed for a better functionality and usage. After performing the preliminary test the acoustical module was improved and three sets of experiments were developed: In the first one, a group of seven exercises with different levels of complexity were carried out. It was found that the blind users were able to externalize the sounds provided by the system with a great accuracy and localize the environmental objects. In the second experiment, the navigation task was analyzed. A scenario based on eight soft objects placed at 2,5m of distance creating a labyrinth was used. During the experiment, remarkable results on object detection and localization were observed, despite simultaneous sounds. Slight errors on navigation accuracy were observed when subjects were navigating through the trajectory. The detection task was successfully carried out; with regards to the localization of the objects, the users perceived small deviations on object lateral localization, i.e., some users had difficulties in detecting the object volume. When forcing the user to pass between the objects, the objective was to localize the objects and to avoid them. The errors appearing in that case could be explained by interference of the reproduction of multiple sounds representing different objects, having the users to precise where the location of each one of the object was. On the third set of experiments the individuals were navigating through controlled and uncontrolled outdoor scenarios (playground of a blind people school, street with crossings, bars, restaurants, parking, kiosks, etc). Despite external noises produced by the environmental objects such as music, people speaking, cars noises, etc…, great results were obtained on object detection and localization, and navigation accuracy. The users were able to avoid all objects and navigate with confidence through such complex environments. 
      In general terms, all these experiments demonstrated that the acoustical representation of the environment is one of the best methods for navigation. It was proven once again that the blind people have considerably great abilities for perceiving the surrounding environment through the hearing. They are able to quickly adapt to the system and to use it as a complementary navigation tool. The acoustical navigation system gives them more confidence and security in navigation. It gives more information from the surrounding environment, information that the white cane is not able to detect. Due to the selected sounds, the system did not interfere with the external noises. Despite all the advantages of the acoustical navigation system, the navigation accuracy depends on the trainings and practice with the use of the device rather than on the sound. All results are based on the end-user feedback giving us directions for refinements, changes, and future work.
      The work developed within the context of this thesis has led to the following publications:

      Journal papers:
      
1. Dunai L., Peris F. G., Garcia B.D., Santiago P. V., Dunai I. (2010) “The influence of the inter-click interval on moving sound source localization for navigation systems”. Applied Physics Journal, 56 (3),  pp. 370-375
2. Dunai L., Peris F. G., Defez B. G., Ortigosa A.N., Brusola S F. (2009). “Perception of the sound source position”, Applied Physics Journal, 55 (3), pp. 448-451 
      
      
      International conference papers:
      
1. Peris F. G., Dunai L., Santiago P. V., Dunai I. (2010). “CASBliP - a new cognitive object detection and orientation aid system for blind people”, CogSys2010 Conference, Zurich 
2. Nuria Ortigosa, Samuel Morillas, Guillermo Peris-Fajarnés and Larisa Dunai. (2010), Disparity maps for free path detection, VISAPP 2010 Conference
3. Dunai L., Peris F G., Defez B. G., Ortigosa A.N., (2009). “Acoustical Navigation System for Visual Impaired People”, LivingAll European Conference 
4. Ortigosa A. N., Dunai L., Peris F. G., Dunai I., Santiago P. V. (2009). “A multiple sensory prototype for visually impaired subject mobility assistance using sound map generation”, LivingAll European Conference
5. Santiago P. V., Ortigosa A.N., Dunai L., Peris F. G., (2009). “Cognitive aid system for blind people (CASbliP)”, INGEGRAF 2009 Conference
6. Ortigosa A. N., Dunai L., Peris. F.G., (2008). Sound map generation for a prototype blind mobility system using multiple sensors”. ABLETECH 08 Conference
7. Fernandes T. M.M., Peris F.G., Dunai L., Redondo J. (2007). “Convolution application in environment sonification for blind people” VII Applied mathematics workshop Valencia
8. Dunai L., Peris F.G., Fernandes T.M.M., Oliver M.J. (2007). “Spatial sound localization base don Fourier Transform”, VII Applied mathematics workshop Valencia
9. Javier Oliver, Alberto Albiol, Guillermo Peris, Larisa Dunai. (2007). “HOG descriptor improvement in person detection by means of the reduction of the space dimensions”., Proceedings of VIII Jornadas de Matemáticas Aplicada, UPV, Spain