The omnipresence of electronic devices in our everyday life goes together with a trend that makes us always more immersed during their utilization. By immersion, we mean that during the development of a new product, it is more and more required to stimulate several senses of the user so as to make the product more attractive. The sense of touch does not escape the rule and is more and more considered. Definitely democratized by its integration in smart phones with touchscreens, the haptic feedback allows enhancing the human-machine interactions in many ways. For instance by improving the comfort of use of a button through the modification of its force feedback. It can also offer an interactive experience during the manipulation of digital information and even improve the communication, particularly through the internet and for blind people, with the introduction of non-verbal signals. For these reasons, the present thesis focuses on the conception of multi-finger haptic devices, a new kind of peripherals integrating multiple actuators and capable of providing a fully programmable force feedback to the user's fingers. A global methodology is presented, outlining the different constituents necessary for their conception: actuator, sensor, control, communication and software user interface. Then, generic tools corresponding to the two first elements are presented. An accurate modeling of miniature electromagnetic short-stroke actuators is made possible thanks to the combination of 3D finite element modeling (FEM) and design of experiments (DOE). The non-usual behavior of magnetic flux lines in miniature actuators with relatively large airgaps imposes to avoid simplified analytical models and to use the reliable results of finite elements. The long computation times required by 3D FEM are balanced by the use of selective DOE making the modeling methodology easily adaptable, rapid and accurate. The parametrical model of the force provided by the modeling methodology is then integrated in a full parametrical setup allowing for the optimization of the actuator force using a conventional algorithm. The advantage of the parametrical optimization is that complementary non-linear constraints such as weight and temperature can be added, making the model multi-physic. Then, several original position measurement techniques using existing sensors are developed including a low-cost custom single-photointerrupter sensor allowing for direction discrimination for fast-prototyping and a hybrid sensing method using tiny Hall sensors and taking advantage of the leaks of the main actuator magnet. Two innovative self-sensing methods are then presented, allowing for the measurement of the mover position of linear short-stroke actuators. The first solution estimates the position of the coil by measuring the acceleration through the back emf. However in this case, a constant acceleration is required, which strongly restrains the application scope. The second solution allows for a real-time measurement of the position thanks to a passive oscillating RLC circuit influenced by the variation of the coil impedance. All the solutions presented are low-cost, compact and require few computation resources. Finally, in order to illustrate the methodology proposed along the thesis, several prototypes are fabricated, giving an overview of the possibilities offered by multi-finger haptic devices. A haptic numeric pad is notably used in an experiment made in collaboration with the University Service of Child and Adolescent Psychiatry in Lausanne with the aim of improving the impaired emotional processing of psychotic adolescents. Moreover, the successful identification of several touch sensations on the same haptic pad lays the first stones of a new tactile language.