In-Depth Notes on Sound Spectrography and Speech Processing

Sound Spectrograph

• A tool to represent sound waves graphically, similar to a knee.

• Reveals common characteristics across different speakers.

• The concept of invariant characteristics allows for effective communication despite variations in vocal timbre.

Invariant Characteristics

• These features exist across various subjects when pronouncing the same words (e.g., "cat").

• They ensure understanding despite differences in individual vocal sounds.

Formants and Formant Transitions

• Formants are segments within the sound spectrograph closely associated with vowels.

• Example: The phrase “Tiny Tim tiptoes with the tulips” illustrates vowel sounds characterized by formants.

• Formant Counts:

• The vowel sound 'a' has three formants.

• The vowel sound 'u' has two formants.

• The distinction in consonants often depends on the transition between formants rather than their individual characteristics.

Consonant Characteristics

• Consonants can be characterized by their formant transitions, seen in the differences of sounds like "sh" and their accompanying spikes in the spectrograph.

• The lowest formant frequency is the first one identified for consonants.

Speech Processing in the Brain

• The brain processes speech signals differently compared to other sounds, supported by infant babbling viewed across many languages.

• The presence of invariant characteristics (formants and transitions) suggests specialized processing in the brain for language comprehension and production.

Categorical Perception

• Introduces the concept of Voicing Onset Time (VOT), which is the delay between the initiation of a sound and the onset of vocal cord vibrations.

• Example Measurements:

• For "ta," VOT is about 91 milliseconds.

• For "da," VOT is around 17 milliseconds.

• Trials involve presenting subjects with varying VOT in random order to identify perceived sounds.

• A critical boundary exists at approximately 40 milliseconds, influencing perception categorically (e.g., distinguishing between "ta" and "da").

Phoneme Processing

• Phonemes are perceived categorically, leading to distinct interpretation ranges, in contrast to other types of sounds that do not follow this pattern.

• This suggests unique cognitive processing characteristics associated with speech sounds.

Neurophysiology and Language

• Studies show that specialized areas of the brain (e.g., Broca's and Wernicke's areas) relate to language processing — Broca's area affects speech production while Wernicke's area impacts comprehension.

• Despite different areas of investigation (infant babbling, categorical perception, invariant characteristics), all support the conclusion that there are specialized mechanisms for processing speech in the brain.