In this paper we discuss the problem of assessing and optimizing speech
enhancement systems. In contrast to speech and audio coding systems,
the speech quality of the input signal as well as of the output signal of a speech enhancement system might be relatively low. Even with unlimited resources it is in general not possible to achieve an output signal with arbitrary high quality. It is emphasized that the basic tradeoff in a speech enhancement system is between the amount of noise reduction and the distortion of the enhanced speech signal. The assessment criteria must be carefully selected to suit the application. A hands-free telephone system serves as an example to show how the criteria "noise reduction" or "echo reduction" and "speech signal distortion" can be used to compare and optimize enhancement systems.