Analysis of French phonetic idiosyncrasies for accent recognition

Speech recognition systems have made tremendous progress since the last few decades. They developed significantly in identifying speech of speaker. However, there is a scope improvement nuances and accents It known that any specific natural language may possess at least one accent. Despite identical word phonemic composition, if it pronounced different accents, we will sound waves, which are from each other. Differences pronunciation, accent intonation general, create most common problems recognition. If lot should acoustic model for separately. We carry out systematic analysis problem accurate classification accents. use traditional machine learning techniques convolutional neural networks, show classical not sufficiently efficient to solve this problem. Using spectrograms signals, propose multi-class framework In paper, focus our attention on French also identify its limitation by understanding impact idiosyncrasies spectrograms.