Objective quality evaluation is a key element for the success of the emerging Voice over IP (VoIP) technologies. Although there are extensive economic incentives for the convergence of voice, data, and video networks, packet networks such as the Internet have inherent incompatibilities with the transport of real time services. Under this paradigm, network planners and administrators are interested in ongoing mechanisms to measure and ensure the quality of these real time services. Objective quality assessment algorithms can be broadly divided into a) intrusive (methods that require a reference signal), and b) non intrusive (methods that do not require a known reference signal). The latter group, typically requires knowledge of the network conditions (level of delay, jitter, packet loss, etc.), and that has been a very active area of research in the past decade. The state of the art methods for objective non-intrusive quality assessment provide high correlations with the subjective tests. Although good correlations have been achieved already for objective non-intrusive quality assessment, the current large voice transport networks are in a hybrid state, where the necessary network parameters cannot easily be observed from the packet traffic between nodes. This thesis proposes a new process, the Network Conditions Estimator (NCE), which can serve as bridge element to real-world hybrid networks. Two classifications systems, an artificial neural network and a C4.5 decision tree, were developed using speech from a database collected from experiments under controlled network conditions. The database was composed of a group of four female speakers and three male speakers, who conducted unscripted conversations without knowledge about the details of the experiment. Using mel frequency cepstral coefficients (MFCCs) as the feature-set, an accuracy of about 70% was achieved in detecting the presence of jitter or packet loss on the channel. This resulting classifier can be incorporated as an input to the E-Model, in order to properly estimate the QoS of a network in real time. Additionally, rather than just providing an estimation of subjective quality of service provided, the NCE provides an insight into the cause for low performance.