Multimedia services, audiovisual applications composed of a combination of discrete and continuous data streams, will be a major part of the traffic flowing in the next generation of high speed networks. The cornerstones for multimedia are Asynchronous Transfer Mode (ATM) foreseen as the technology for the future Broadband Integrated Services Digital Network (B-ISDN) and audio and video compression algorithms such as MPEG-2 that reduce applications bandwidth requirements. Powerful desktop computers available today can integrate seamlessly the network access and the applications and thus bring the new multimedia services to home and business users. Among these services, those based on multipoint capabilities are expected to play a major role. Interactive multimedia applications unlike traditional data transfer applications have stringent simultaneous requirements in terms of loss and delay jitter due to the nature of audiovisual information. In addition, such stream-based applications deliver data at a variable rate, in particular if a constant quality is required. ATM, is able to integrate traffic of different nature within a single network creating interactions of different types that translate into delay jitter and loss. Traditional protocol layers do not have the appropriate mechanisms to provide the required network quality of service (QoS) for such interactive variable bit rate (VBR) multimedia multipoint applications. This lack of functionalities calls for the design of protocol layers with the appropriate functions to handle the stringent requirements of multimedia. This thesis contributes to the solution of this problem by proposing new Network Adaptation and ATM Adaptation Layers for interactive VBR multimedia multipoint services. The foundations to build these new multimedia protocol layers are twofold; the requirements of real-time multimedia applications and the nature of compressed audiovisual data. On this basis, we present a set of design principles we consider as mandatory for a generic Multimedia AAL capable of handling interactive VBR multimedia applications in point-to-point as well as multicast environments. These design principles are then used as a foundation to derive a first set of functions for the MAAL, namely; cell loss detection via sequence numbering, packet delineation, dummy cell insertion and cell loss correction via RSE FEC techniques. The proposed functions, partly based on some theoretical studies, are implemented and evaluated in a simulated environment. Performances are evaluated from the network point of view using classic metrics such as cell and packet loss. We also study the behavior of the cell loss process in order to evaluate the efficiency to be expected from the proposed cell loss correction method. We also discuss the difficulties to map network QoS parameters to user QoS parameters for multimedia applications and especially for video information. In order to present a complete performance evaluation that is also meaningful to the end-user, we make use of the MPQM metric to map the obtained network performance results to a user level. We evaluate the impact that cell loss has onto video and also the improvements achieved with the MAAL. All performance results are compared to an equivalent implementation based on AAL5, as specified by the current ITU-T and ATM Forum standards. An AAL has to be by definition generic. But to fully exploit the functionalities of the AAL layer, it is necessary to have a protocol layer that will efficiently interface the network and the applications. This role is devoted to the Network Adaptation Layer. The network adaptation layer (NAL) we propose, aims at efficiently interface the applications to the underlying network to achieve a reliable but low overhead transmission of video streams. Since this requires an a priori knowledge of the information structure to be transmitted, we propose the NAL to be codec specific. The NAL targets interactive multimedia applications. These applications share a set of common requirements independent of the encoding scheme used. This calls for the definition of a set of design principles that should be shared by any NAL even if the implementation of the functions themselves is codec specific. On the basis of the design principles, we derive the common functions that NALs have to perform which are mainly two; the segmentation and reassembly of data packets and the selective data protection. On this basis, we develop an MPEG-2 specific NAL. It provides a perceptual syntactic information protection, the PSIP, which results in an intelligent and minimum overhead protection of video information. The PSIP takes advantage of the hierarchical organization of the compressed video data, common to the majority of the compression algorithms, to perform a selective data protection based on the perceptual relevance of the syntactic information. The transmission over the combined NAL-MAAL layers shows significant improvement in terms of CLR and perceptual quality compared to equivalent transmissions over AAL5 with the same overhead. The usage of the MPQM as a performance metric, which is one of the main contributions of this thesis, leads to a very interesting observation. The experimental results show that for unexpectedly high CLRs, the average perceptual quality remains close to the original value. The economical potential of such an observation is very important. Given that the data flows are VBR, it is possible to improve network utilization by means of statistical multiplexing. It is therefore possible to reduce the cost per communication by increasing the number of connections with a minimal loss in quality. This conclusion could not have been derived without the combined usage of perceptual and network QoS metrics, which have been able to unveil the economic potential of perceptually protected streams. The proposed concepts are finally tested in a real environment where a proof-of-concept implementation of the MAAL has shown a behavior close to the simulated results therefore validating the proposed multimedia protocol layers.