Origins of scale invariance in vocalization sequences and speech

Abstract

To communicate effectively animals need to detect temporal vocalization cues that vary over several orders of magnitude in their amplitude and frequency content. This large range of temporal cues is evident in the power-law scale-invariant relationship between the power of temporal fluctuations in sounds and the sound modulation frequency (f). Though various forms of scale invariance have been described for natural sounds, the origins and implications of scale invariant phenomenon remain unknown. Using animal vocalization sequences, including continuous human speech, and a stochastic model of temporal amplitude fluctuations we demonstrate that temporal acoustic edges are the primary acoustic cue accounting for the scale invariant phenomenon. The modulation spectrum of vocalization sequences and the model both exhibit a dual regime lowpass structure with a flat region at low modulation frequencies and scale invariant 1/f2 trend for high modulation frequencies. Moreover, we find a time-frequency tradeoff between the average vocalization duration of each vocalization sequence and the cutoff frequency beyond which scale invariant behavior is observed. These results indicate that temporal edges are universal features responsible for scale invariance in vocalized sounds. This is significant since temporal acoustic edges are salient perceptually and the auditory system could exploit such statistical regularities to minimize redundancies and generate compact neural representations of vocalized sounds.

A Neural Ensemble Correlation Code for Sound Category Identification

“…By comparison, vocalized sounds have a somewhat lower nonstationary index (SI=0.28 ±0.04; mean ±SEM) at the analysis time-scales employed. This is expected given that vocalizations transition from periods of silence and vocalizations at time-scales of just a few Hz 26 and even within vocalization segments the correlation structure can vary dynamically from moment-to-moment (e.g., Fig. 5).…”

“…This contrasts the running water excerpt, where the correlations are largely diagonalized, indicating minimal correlation between distant frequency channels. The temporal correlation structure in speech exhibits a relatively slow temporal structure (64 ms half width), which reflects the relatively slow time-varying structure of speech elements and words 25,26 . By comparison, the water sound has a relatively fast temporal correlation structure (4 ms half width) that is indicative of substantially faster temporal fluctuations in the sound power.…”

Neural timing of stimulus events with microsecond precision

“…Precise timing of neural signals could encode rich information that is relevant to a variety of brain functions [ 73 ]. For example, the temporal precision of neural events may be a key property in categorizing sound features by the auditory system [ 74 , 75 ]. Indeed, the information theoretic analysis showed that EFPs from many recording sites in the bat IC could unambiguously differentiate three out of four simulated echolocation calls of different durations.…”

Spiking network optimized for word recognition in noise predicts auditory system hierarchy

“…As seen from the optimal HSNN receptive fields and observed physiologically (Fig 4), the early levels of the HSNN and auditory pathway operate on and encode fine temporal details and these transition to a coarser temporal representation. The deep layers of the network largely operate on relatively slow features in speech, such as consonant vowel transitions, syllable structure, and time-varying formant structure, which are known to be critical for recognition task [31,33,53]. The early levels of the optimal HSNN synchronize and preserves envelopes in speech exceeding several hundredths of Hz and this synchronization ability is sequentially degraded across layers (Fig 7C), analogous to changes in synchronization ability seen along the auditory pathway [5].…”

“…We focused on the fluctuation / rhythm band (1-25 Hz) and the periodicity pitch (75-150 Hz) bands because of their perceptual relevance and because neurons at various level of the auditory pathway can synchronize to the temporal envelopes for these bands. The fluctuation band is particularly relevant for speech recognition and contains much of the low modulation frequency information content required for word identification [5,31,33]. By comparison, the periodicity pitch band encompasses the temporal fluctuations created by vocal fold vibration which can extend out to several hundred Hz [5,33].…”