The research reported in this thesis proposes, designs and tests the Quality-Oriented Adaptation Scheme (QOAS), an application-level adaptive scheme that offers high quality multimedia services to home residences and business premises via local broadband IP-networks in the presence of other traffic of different types. QOAS uses a novel client-located grading scheme that maps some network-related parameters’ values, variations and variation patterns (e.g. delay, jitter, loss rate) to application-level scores that describe the quality of delivery. This grading scheme also involves an objective metric that estimates the end-user perceived quality, increasing its effectiveness. A server-located arbiter takes content and rate adaptation decisions based on these quality scores, which is the only information sent via feedback by the clients. QOAS has been modelled, implemented and tested through simulations and an instantiation of it has been realized in a prototype system. The performance was assessed in terms of estimated end-user perceived quality, network utilisation, loss rate and number of customers served by a fixed infrastructure. The influence of variations in the parameters used by QOAS and of the networkrelated characteristics was studied. The scheme’s adaptive reaction was tested with background traffic of different type, size and variation patterns and in the presence of concurrent multimedia streaming processes subject to user-interactions. The results show that the performance of QOAS was very close to that of an ideal adaptive scheme. In comparison with other adaptive schemes QOAS allows for a significant increase in the number of simultaneous users while maintaining a good end-user perceived quality. These results are verified by a set of subjective tests that have been performed on viewers using a prototype system.