Effects of Noise on the Behavioral and Neural Categorization of Speech

Abstract

We investigated whether the categorical perception (CP) of speech might also provide a mechanism that aids its perception in noise. We varied signal-to-noise ratio (SNR) [clear, 0 dB, −5 dB] while listeners classified an acoustic-phonetic continuum (/u/ to /a/). Noise-related changes in behavioral categorization were only observed at the lowest SNR. Event-related brain potentials (ERPs) differentiated category vs. category-ambiguous speech by the P2 wave (˜180-320 ms). Paralleling behavior, neural responses to speech with clear phonetic status (i.e., continuum endpoints) were robust to noise down to −5 dB SNR, whereas responses to ambiguous tokens declined with decreasing SNR. Results demonstrate that phonetic speech representations are more resistant to degradation than corresponding acoustic representations. Findings suggest the mere process of binning speech sounds into categories provides a robust mechanism to aid figure-ground speech perception by fortifying abstract categories from the acoustic signal and making the speech code more resistant to external interferences.

Lexical influences on categorical speech perception are driven by a temporoparietal circuit

“…Adaptation studies-in which continuum sounds are presented repetitively and/or in serial order (Eimas and Corbit, 1973;Miller, 1975 causing a boundary shift in the direction toward the un-adapted detector at the polar end of the continuum (Rozsypal et al, 1985). As confirmed empirically, larger boundary shifts would be expected for less strongly categorized continua (Rozsypal et al, 1985), e.g., vowels vs. stop consonants (Altmann et al, 2014), acoustically degraded speech (Bidelman et al, 2020;, and for ambiguous speech tokens as shown here and previously (Ganong, 1980;Gow et al, 2008;Lam et al, 2017;Myers and Blumstein, 2008;Noe and Fischer-Baum, 2020).…”

Auditory cortex is susceptible to lexical influence as revealed by informational vs. energetic masking of speech categorization

“…The introduction of acoustic degradation (i.e., noise) prolongs and weakens the ERPs, indicating predictable masking on speech processing (Bidelman and Howell, 2016;Bidelman et al, 2018;Billings et al, 2009). However, we have recently shown that speech categories-those carrying a strong phonetic identity-are more resilient to noise degradation than their phonetically ambiguous counterparts (Bidelman et al, 2020a). Categorization recruits a wide variety of frontal, temporal, and parietal brain regions (Al-Fahad et al, 2020;Chang et al, 2010;Myers et al, 2009).…”

Data-driven machine learning models for decoding speech categorization from evoked brain responses

“…Forty-nine young adults (male: 15, female: 34; aged 18 to 33 years) were recruited as participants from the University of Memphis student body to participate into our ongoing studies on the neural basis of speech perception and auditory categorization (Bidelman et al, 2020b;Bidelman & Walker, 2017b;Mankel et al, 2020). All participants had normal hearing sensitivity (i.e., <25 dB HL between 500-2000 Hz).…”

“…Since we were interested to decode prototypical (Tk1/5) from ambiguous speech (Tk3)a marker of categorical processing (Bidelman, 2015;Bidelman & Walker, 2019b;Liebenthal et 10 al., 2010b)-we merged Tk1 and Tk5 responses since they reflect prototypical vowel categories ("u" vs. "a'). In contrast, Tk3 reflects a bistable percept-an category-ambiguous sound listeners sometimes label as "u" or "a" (Bidelman et al, 2020a;Bidelman & Walker, 2017b;Mankel et al, 2020). To ensure an equal number of trials and signal to noise ratio (SNR) for prototypical and ambiguous stimuli, we considered only 50% of the data from the merged (Tk1/5) samples.…”

“…In particular, several studies have shown that the efficiency of listeners' speech categorization varies in accordance with their underlying brain activity (Bidelman et al, 2013a;Bidelman & Alain, 2015;Bidelman & Lee, 2015b;Perlovsky, 2011). For example, Bidelman et al demonstrated that brain responses in the time frame of 180-320 ms were more robust for phonetic prototypes vs. ambiguous speech tokens, thereby reflecting category-level processing (Bidelman et al, 2020a). Other studies have shown links between N1-P2 amplitudes of the auditory cortical ERPs and the strength of listeners' speech identification (Bidelman & Walker, 2017a) and labeling speeds (Al-Fahad et al, 2020) in speech categorization tasks (Bidelman et al, 2014;Bidelman & Alain, 2015).…”