Lecture 2 Notes: Frequency Selectivity and Speech Perception

Frequency Selectivity and Speech Perception

Objectives

  • Explain frequency selectivity.
  • Explain the importance of frequency selectivity for speech perception using vowels.
  • Understand hearing as a combination of bottom-up and top-down mechanisms.

Demonstrations of Frequency Selectivity

  • Listeners find it easier to differentiate instruments (cello and flute) playing together compared to two cellos, because the flute and cello excite different places on the basilar membrane, enhancing frequency selectivity.

Frequency Selectivity Definition

  • Frequency selectivity is the ability to separate sounds with different frequencies occurring simultaneously.

Basilar Membrane and Frequency Selectivity

  • The basilar membrane's selectivity is demonstrated by its velocity at a tested location depending on the frequency of the tone; this is measured using Laser Interferometry, which measures the speed of the basilar membrane's movement.
  • Frequency selectivity interacts with sound intensity; higher intensities correspond to lower frequencies (e.g., 7 kHz), and lower intensities correspond to higher frequencies (e.g., 10 kHz).

Tuning in the Auditory Nerve (Evans, 1975)

  • Each auditory-nerve fibre responds to a narrow range of frequencies, demonstrated using micro-electrode recordings and tuning curves.
  • Frequency specificity reduces and shifts as a function of intensity.

Psychoacoustic Demonstrations of Frequency Selectivity

  • Warbling tones centered on 500 Hz, 1000 Hz, and 2000 Hz.
  • A band of noise centered on 1000 Hz can mask a tone, showing frequency selectivity.
  • Tones at 500 Hz and 2000 Hz are resolved from noise, whereas a tone at 1000 Hz is not.

Psychophysical Tuning Curves

  • The level of a masker tone is adjusted until it prevents the listener from detecting a test tone. This is done for many different masker tone frequencies.
  • The closer the masker frequency is to the test tone, the lower the masker level needed to mask the test tone.
  • This indicates frequency selectivity.
  • Example: Test tone = 4 kHz.

Interim Summary

  • Frequency selectivity is the ability to separate energy at different frequencies.
  • It is determined by the properties of the basilar membrane.
  • Tuning is observed on the basilar membrane (laser interferometry), in the fibres of the auditory nerve (single-unit recordings), and in the whole system (psychophysically).

Importance of Frequency Selectivity

  • Crucial for speech perception.

Spectrogram Analysis of Speech

  • Spectrogram displays frequency vs. time, showing speech sounds such as fricatives (s, sh, z).
  • Formants in vowels are visible, with examples provided.

Vowel Production

  • Vowel sounds ("ee," "ar," "oo") are produced with specific resonances of the vocal tract and radiated from the lips. Spectrum of resonances of vocal tract and radiated from lips (first three formants)
  • The vocal tract profile influences vowel sounds.

Vowel Spectra

  • Vowels have distinct spectra with discernible formants F1F1 and F2F2.
  • Sounds synthesized from harmonics, with pairs indicating F1F1 and F2F2, can recreate recognizable vowel sounds.

Interim Summary

  • Vowels are characterized by peaks and valleys in their spectra.
  • Peaks correspond to vocal tract resonances called formants.
  • The frequencies of the first two formants are the primary cues distinguishing vowels in English.
  • Simplified vowel spectra containing just two pairs of harmonics can still be identified as the vowels they model.

Top-Down Effects on Hearing

  • Building on the bottom-up approach (basilar membrane, frequency selectivity).
  • Language knowledge provides top-down support for phoneme and word identification.
  • Illustrations include the Ganong effect, phoneme restoration, sound and word reduction, the McGurk effect, and sinewave speech.

The Auditory Nervous System

  • Bottom-up connections: Cochlea → Cochlear nucleus → Superior Olive → Inferior colliculus → Auditory Cortex

Ganong Effect (1980)

  • Ambiguous phonemes are perceived differently based on context.
  • Examples: "giss" vs. "kiss," "gift" vs. "kift."
  • Lexical knowledge influences perception.

Phoneme Restoration Effect (Warren, 1970)

  • Missing phonemes are restored based on context.
  • Example: "It was found the *eel was on the table" (heard as "meal"), "It was found the *eel was on the shoe" (heard as "heel").
  • Semantic knowledge influences perception.

Sound and Word Reduction

  • In connected speech, words are often acoustically reduced.
  • Syntactic and semantic knowledge aids in understanding.

McGurk Effect (McGurk & MacDonald, 1976)

  • Visual cues (lip movements) influence auditory perception.
  • Example: Seeing "fa" while hearing "ba" can result in perceiving "fa."
  • Articulatory knowledge is involved.

Sinewave Speech

  • Speech signal in which formants have been replaced with pure tones tracking intensity modulations.
  • Example: "The dessert was put in the oven at the start of the meal."
  • Phonological, lexical, syntactic, and semantic knowledge enable understanding.

Top-Down Effects in the Brain

  • Top-down connections: Auditory Cortex → Inferior colliculus → Superior Olive → Cochlear nucleus → Cochlea
  • Efferent connections exist throughout the auditory pathway.
  • Top-down effects mostly occur at the cortical level, involving associative areas in addition to the primary auditory cortex.
  • Little evidence supports top-down effects on lower-level neural regions (brainstem and cochlea).

Selective Attention and Brainstem Activity

  • Selective attention during speech perception in noise affects brainstem activity (inattentional deafness).
  • Inattentional deafness: Reduced sensitivity to auditory stimuli when attention is engaged in another task.
  • The auditory-evoked brainstem response decreases as working memory load increases (Sörqvist, Stenfelt, & Rönnberg, 2012).

Cochlear Tuning via Superior Olive

  • Evidence suggests top-down connections from the superior olive to the cochlea during selective attention.
  • These connections sharpen the responsiveness of inner hair cells, helping detect tones in noise by enhancing frequency selectivity.

Medial Olivo-Cochlear Reflex

  • Control loop between the cochlea and superior olive reduces cochlear responsiveness to sudden loud noises.

Top-Down Effects: Summary

  • Hearing involves both bottom-up and top-down processes.
  • Lexical and contextual effects facilitate quick and effective language comprehension at the cortical level.
  • Top-down effects to sub-cortical and cochlear levels are observed during selective attention (speech in noise).

Summary

  • Frequency selectivity is the cochlea's ability to respond selectively to different frequencies, acting as a frequency analyser.
  • It allows us to identify frequencies defining vowel identity (formants).
  • Hearing combines bottom-up (signal-driven) and top-down (knowledge-driven) processes.
  • Linguistic top-down processes fill signal gaps, correct imperfections, and anticipate upcoming information cortically.
  • Selective attention involves top-down sub-cortical connections affecting brainstem functions and cochlear tuning.