Hands-free operation of speech processing systems is sometimes desired to avoid encumbrance of the user by tethered microphone equipment. This paper explores the use of array microphones and neural networks (MANN) for robust speech recognition in real-world environments, such as large-group conferencing. Microphone arrays (MA) provide high-quality, hands-free sound pickup under severe acoustical conditions; and neural network (NN) processors "learn" the characteristics of environmental interference and transform features of MA-enhanced signal to those obtained under close-talking conditions. In this study, both realroom collected and computer-simulated reverberant speech signals are used to evaluate the power and advantages of MANN for direct deployment of speech recognition technology in adverse practical environments.