LATENT LIBRARY DEEP LEARNING Speech andAudio AI Recognition, Synthesis and SoundUnderstanding LATENT LIBRARY EDITORIAL

Speech and Audio AI

Recognition, Synthesis and Sound Understanding

Latent Library Editorial

IntroductionIntroduction

Voice is the most natural interface humans have. We speak before we read, and we recognise a friend's voice, a crying baby or a failing engine instantly. Computers have long struggled with sound, but in recent years speech recognition became accurate enough for everyday dictation and captioning, synthetic voices became hard to tell from real ones, and audio models learned to recognise everything from bird songs to machine faults.

This book explains how machines hear and speak. It starts from first principles (what sound is, how it is recorded and how it is turned into pictures called spectrograms) and builds up to speech recognition, speech synthesis, speaker recognition, sound classification, voice assistants and translation. Throughout, it pays attention to what goes wrong: accents and noise, latency, privacy, and the serious ethical questions raised by voice cloning.

No signal-processing background is assumed.

How the book is organised

Chapter 1Why Audio AI Matters

Audio AI is already everywhere:

Each application has different demands for accuracy, speed, robustness and privacy.

You have reached the end of the free sample.
The full book continues with 43 more chapters, plus the appendices.

Get the full book