Towards Automatic Analysis of Audio Recordings from Children with Autism Spectrum Disorder