Adattamento di una rete ricorrente per l'analisi tonale di file audio

New developments in artificial intelligence, machine learning and neural networks are offering advanced solutions in many computer science fields. Thanks to these developments, performances over speech recognition tasks has markedly increased in the last decade, allowing researchers to solve complex problems in this context. Consequences of these improvements involve a field that has some elements in common with speech recognition: tonal analysis. Speech recognition tasks and tonal analysis share similar input and output: an audio containing information we want to comprehend as input, and a transcript of these information as output. The main purpose of this work is to adapt to tonal analysis a set of techniques, methodologies and models related to speech recognition, in order to predict sequences of chords played inside audio tracks.