This paper deals with the application of information extracted from AM and FM time-frequency representations of speech to the task of determining speech quality. The representations are introduced and then the procedure for data extraction is outlined. The experimental setup for the assessment of objective quality covers distortions typically found in speech communication systems. To determine how well these quality measures perform, regression analysis is used to evaluate how well they estimate the results of subjective testing. Considering each class of distortions individually the objective measures demonstrate good performance, however, this level does not seem to hold as well in the aggregate case. This leads to suggestions as to where possible improvements can be made to the procedure.