Before jumping to conclusions about this being the same as dithering or any other conclusions, it would be good to read the whole research paper, and not just the headlines.
This is a new finding, especially the research into the cause of what is happening.
Dithering is a technique to add random noise so that things don’t "stick" to a given value, typically in a digitized system, but not exclusively. Let’s say you are trying to digitize a value, and that value if you have the resolution was 12.7, but since you don’t, you get 12 (or 13) and it never changes. Adding dithering will make that 12.7 bounce between 12 and 13 (or higher) such that the average = 12.7.
In the paper (research done on mice), what they found is that the white noise (fairly high levels) suppressed neural activity. That would be the opposite of adding "bias" .. more like an automatic gain control turning down the gain. The white noise reduced the firing rate of the neurons associated with the frequencies, but increased the accuracy. The white noise was within 10db of the signal. Suppression of neurons would not be akin to dithering either.
Noise would typically impair frequency discrimination and impairs most auditory tasks. What this study showed was that noise could improve frequency discrimination when the frequencies are close to each other. It showed no improvement when they are farther apart.
This research paper would be more inline with what you suggest in your OP, w.r.t. low level noise increasing perception: https://www.ncbi.nlm.nih.gov/pubmed/19558607?dopt=Abstract , though they did not dig as deep into the underlying neural responses. While in the paper posted above, the neural responses were repressed, in this paper, they were increased. I could see that be indicative of dithering since neurons can be threshold triggered, more indicative of digital response than analog.
... and jnorris2005, +1 !!
This is a new finding, especially the research into the cause of what is happening.
Dithering is a technique to add random noise so that things don’t "stick" to a given value, typically in a digitized system, but not exclusively. Let’s say you are trying to digitize a value, and that value if you have the resolution was 12.7, but since you don’t, you get 12 (or 13) and it never changes. Adding dithering will make that 12.7 bounce between 12 and 13 (or higher) such that the average = 12.7.
In the paper (research done on mice), what they found is that the white noise (fairly high levels) suppressed neural activity. That would be the opposite of adding "bias" .. more like an automatic gain control turning down the gain. The white noise reduced the firing rate of the neurons associated with the frequencies, but increased the accuracy. The white noise was within 10db of the signal. Suppression of neurons would not be akin to dithering either.
Noise would typically impair frequency discrimination and impairs most auditory tasks. What this study showed was that noise could improve frequency discrimination when the frequencies are close to each other. It showed no improvement when they are farther apart.
This research paper would be more inline with what you suggest in your OP, w.r.t. low level noise increasing perception: https://www.ncbi.nlm.nih.gov/pubmed/19558607?dopt=Abstract , though they did not dig as deep into the underlying neural responses. While in the paper posted above, the neural responses were repressed, in this paper, they were increased. I could see that be indicative of dithering since neurons can be threshold triggered, more indicative of digital response than analog.
... and jnorris2005, +1 !!