Would you be writing your own Speech analyser or using an off the shelf product and customising it?