Speech rate measured in bit/sec

Are there references giving the speech rate measured in bit/sec?

I easily found references on speech rate giving measurement in phones/sec, syllables/sec or words/sec. Are there references giving the speech rate in bit/sec, i.e., answering the question how much information we get across while talking to each other?

I am both interested in average values for normal speech and peak values for fast speech.

EDIT: Maybe the answer is "There are no such references". I am also willing to accept a reasonably high level answer of this kind.

• How would you measure information 'bits' -- in phonemes? Nov 28, 2016 at 16:13
• @StoneyB I leave the derivation of bits to the authors of the references. Phonemes (using the entropy of the phoneme frequency distribution) are clearly a good starter, but other ideas are possible and equally wellcome. Note that this is a [[:tag:reference-request]], not an invitation to do ad hoc calculations. Nov 29, 2016 at 9:37
• Can we assume you don't mean bit-rate for sampled speech that yields an intelligible signal? Nov 30, 2016 at 16:38
• @user6726: I am tempted to say Yes, I don't mean that. But maybe one can do something interesting with it: What is the lowest sampling bit rate that still yields an intelligible signal? This kind of arg min may be meaningful. Nov 30, 2016 at 16:52
• This paper might be helpful: scitation.aip.org/content/asa/journal/jasa/103/5/10.1121/…. Nov 30, 2016 at 17:43

It looks like Pierce & Karlin 1957 might be a good starting point. It came out almost a decade after the original Shannon (1948) entropy paper and a decade before Kolmogorov's 1968 paper on quantifying information. Here's the abstract:

The limitation on the rate at which information can be transmitted over an ordinary telephone channel is a human one. In this study people read words as fast as they were able to; from these results some deductions are made about the capacity of a human being as an information channel. The discrepancy between human channel capacity measured thus (40-50 bits/sec) and telephone and television channel capacity (about 50,000 bits/sec and 50,000,000 bits/ sec respectively) is provocative.

Also, I feel like Florian Jaeger has flirted with this unit of measurement when dealing with information density, but I can't remember an article name to save my life.

• Can you please edit this post to add a summary of the paper's results so that it actually answers the question? :) Dec 1, 2016 at 14:49
• Having a number from Jaeger would be gorgeous, but he navigates around the question of the channel capacity cautiously, e.g., in this 2010 paper he uses an arbitrary value of 4 bit/word just for illustration: ac.els-cdn.com/S0010028510000083/… (T. Florian Jaeger, Redundancy and reduction: Speakers manage syntactic information density, Cognitive Psychology **61**(2010) 23–62) Dec 8, 2016 at 10:57

I can't pretend to really understand this article, but the authors have an algorithm for encoding speech at 800 bps (as they say, using split vector quantization). They have tested it using male and female voices and seem satisfied that the output is comparable to standard 2400 bps LPC-10 encoding, though the authors are electrical engineers, not psychologists, so there isn't a body of perceptual studies to go with the method. On the other hand, this study does have a perceptual study to go with the technology, and an algorithmic tweak (based on a theory of auditory perception) results in improves perceptual quality for 2400 bps LPC coding. Chapter 2 of the book Ultra Low Bit-Rate Speech Coding seems to be out there (here). It appears (and I must emphasize "appears") that one can reach 120 bps coding. It might be worth posting a related question to Signal Processing SE, to see especially to what extent these ridiculously low rates yield intelligible speech (esp. to get a technically-qualified evaluation of the state of the art).

This thesis suggests 20 bits/sec, based on Shannon's estimate of 1 bit per character entropy in English text, 20 phonemes/second speech rate, and about 1 character/phoneme.

• The number 20 bit/sec from the thesis added is on the low side and more like an ad hoc estimate. It uses the rather crude approximation, that one phoneme is represented by one letter in English. Dec 8, 2016 at 11:00

I finally came across a paper where the authors state a speech rate in bit/s. The paper is Christophe Coupé, Yoon Oh, Dan Dediu, and François Pellegrino, Different languages, similar encoding efficiency: Comparable information rates across the human communicative niche and they give a value of `39 ± 5 bit/s`.