Neural Sync Active

Learning Objectives

Registry Synced

Learning Objectives

388 words

2 min read

Reading compass

Now · 1. Why Speech?

Learning Objectives

Understand speech production and perception
Describe the source-filter model of speech
Identify key applications of speech technology
Basic signal processing concepts
Understanding of frequency and time domains

1. Why Speech?

Speech is the most natural form of human communication. Speech technology enables machines to understand, synthesize, and respond to human speech. Applications: Voice assistants (Siri, Alexa), automatic transcription, speaker verification, text-to-speech, hearing aids, language learning.

2. Speech Production

The human speech production system:

Lungs: Provide airflow (power source)
Vocal folds (larynx): Vibrate to produce voiced sounds
Vocal tract (pharynx, mouth, nasal cavity): Filters sound, creates resonances Source-Filter Model:

Source: Glottal waveform (periodic for voiced, noise-like for unvoiced)
Filter: Vocal tract transfer function (shapes the spectrum)
Output: Speech signal = Source * Filter (convolution in time, multiplication in frequency)

3. Speech Sounds

Phonemes: Basic units of sound. English has ~44 phonemes. Voiced vs Unvoiced: Voiced = vocal folds vibrate (vowels, b, d, g). Unvoiced = no vibration (p, t, k, s, f). Place and Manner of Articulation: Where and how the sound is produced (lips, tongue, palate, glottis).

Q1: What is the source-filter model of speech production?

Speech = Source signal (glottal waveform) filtered by vocal tract. Source provides excitation (periodic for voiced, noise for unvoiced). Filter (vocal tract shape) determines spectral envelope (formants). Q2: What is the difference between voiced and unvoiced sounds?

Voiced: vocal folds vibrate, creating periodic signal (vowels, nasals). Unvoiced: vocal folds open, turbulent airflow creates noise-like signal (fricatives: s, f, sh). Q3: What are formants?

Resonant frequencies of the vocal tract. First two formants (F1, F2) distinguish vowels. F1 correlates with jaw opening, F2 with tongue position. Formants are visible as dark bands on spectrograms. Q4: What is the difference between phonemes and phones?

Phonemes: abstract sound categories that distinguish meaning (English /p/ and /b/ are different phonemes). Phones: actual physical sounds produced. Allophones: different phones representing same phoneme (aspirated p in "pin" vs unaspirated in "spin"). Q5: Why is speech recognition hard?

Variability: speaker differences (age, gender, accent), speaking rate, coarticulation (sounds influenced by neighbors), background noise, homophones (same sound, different words), ambiguity (segmentation, context). Join Discord NextDigital Signal Processing Fundamentals