In this chapter, we'll deal with sounds and speech. Sound data comes in the form of waves, and therefore requires different preprocessing than other types of data.

Machine learning on audio signals finds commercial applications in speech enhancement (for example, in hearing aids), speech-to-text and text-to-speech, noise cancellation (as in headphones), recommending music to users based on their preferences (such as Spotify), and generating audio. Many fun problems can be encountered in audio, including the classification of music genres, the transcription of music, generating music, and many more besides.

We'll implement several applications with sound and speech in this chapter. We'll first do a simple example of a classification task, where we try to distinguish different words. This would be a typical application in a smart...