In this paper, we review several promising methods that were proposed in the past few years to deal with the problem of speech recognition in adverse conditions. We discuss these methods in six categories: signal enhancement preprocessing, special transducer arrangements, noise masking, stress compensation, robust distortion measures, and novel speech representations. We explain each categorical approach and provide a digest of the performance improvements each method is able to achieve. This type of information is helpful in making a technical decision for the actual recognizer design to be deployed in adverse environments.