Speech earthquakes:

Speech earthquakes: scaling and universality in human voice

Jordi Luque, Bartolo Luque, Lucas Lacasa

(Submitted on 5 Aug 2014)

http://arxiv.org/pdf/1408.0985.pdf

Speech is a distinctive 

complex feature of human capabilities. 

In order to understand the physics 

underlying speech production, 

in this work we empirically analyse 

the statistics of large human speech datasets 

ranging several languages. 

We first show that during speech 

the energy is unevenly released 

and power-law distributed, 

reporting a universal robust 

Gutenberg-Richter-like law in speech. 

We further show 

that such earthquakes in speech 

show temporal correlations, 

as the interevent statistics 

are again power-law distributed. 

Since this feature takes place 

in the intra-phoneme range,

we conjecture that the responsible 

for this complex phenomenon 

is not cognitive, but it resides on the 

physiological speech production mechanism. 

Moreover, we show 

that these waiting time distributions 

are scale invariant under 

a renormalisation group transformation, 

suggesting that the process of speech generation 

is indeed operating close to a critical point. 

These results are put in contrast 

with current paradigms in speech processing, 

which point towards low dimensional deterministic chaos 

as the origin of nonlinear traits in speech fluctuations. 

As these latter fluctuations are indeed 

the aspects that humanize synthetic speech, 

these findings may have an impact 

in future speech synthesis technologies. 

Results are robust and independent 

of the communication language 

or the number of speakers, 

pointing towards an universal pattern 

and yet another hint 

of complexity in human speech.

Comments: Submitted for publication

Subjects: Physics and Society (physics.soc-ph); Computation and Language (cs.CL); Neurons and Cognition (q-bio.NC)

Cite as: arXiv:1408.0985 [physics.soc-ph]

  (or arXiv:1408.0985v1 [physics.soc-ph] for this version)