Recent developments in speech recognition under adverse conditions