Automatic speech recognition systems achieve good performance when they have to transcribe prepared speech (close to read text) but have real difficulties to deal with spontaneous speech (non-prepared speech, conversational speech). Our work focused on the new challenges brought by this spontaneity of the speech, making it dfficult to be transcribed by the existing automatic speech recognition systems. We studied how to improve global performance of automatic speech recognition systems towards spontaneous speech by adapting language model and pronunciation dictionary to this particular type of speech. We also studied the detection of disuent speech portions (produced by spontaneous speech) in speech signal using a Gaussian Mixture Model (GMM)-based classffier trained on prosodic features covering the main prosodic characteristics (duration, fundamental frequency and energy).