To build systems that understand speech, we must first understand the structure of speech itself and how it is represented digitally. This chapter provides the necessary background, starting from the properties of a sound wave and ending with its numerical representation on a computer, ready for processing.

We will cover the complete path from a spoken utterance to a machine-readable format. You will learn about the fundamental properties of human speech and the technical steps required to digitize it. A continuous analog audio signal, represented as $x(t)$, must be converted into a discrete sequence of numbers, $x[n]$, through processes like sampling and quantization.

By the end of this chapter, you will be able to:

• Describe the high-level architecture of a typical Automatic Speech Recognition (ASR) system.
• Identify the basic linguistic units of speech, such as phonemes and allophones.
• Explain how analog audio is converted to digital signals through sampling and quantization.
• Use the Python Librosa library to load and manipulate audio data.
• Distinguish between time-domain and frequency-domain representations of a signal.
• Generate and interpret spectrograms as a way to visualize speech.

The chapter concludes with a hands-on exercise where you will apply these skills to load and visualize audio waveforms and spectrograms, setting the stage for the feature extraction methods that follow.