Matthew Diamond and Tobias Heed started off with a lovely summary of sensory perception in skin and whiskers. They emphasized the importance of temporal aspects of touch perception. At the base of each ridge of fingerprint, there are sensors for strain and pressure(?) that provide high temporal bandwidth signals that are essential for perception of surface roughness and stickiness. Under carefully controlled conditions, the skin and whisker (in mouse/rat) touch sensors show spike jitter of under 100us.
Next Michael Schmuker talked about olfaction. He started off pointing out that half of taste is smell, and that entire businesses such as hygiene products and airport perfume sales depend on human smell.
He pointed out the Jim Bower ;pointed out in 2013 that we now have the tools to better understand olfaction with multi-unit recordings, but we cannot record or play back smells. In chemistry, we have molecules, but they are discrete, not continuous in identity, since they consist of different hydrocarbon chain compounds. The space is high dimensional. Oderants have up to 7-8 or even 13 heavy atoms in them (C or O) and there are a huge number of permutations of these.
Receptors in nose have very broad sensitivities. A particular receptor might have 1 preferred ligand, but others overlap its receptive field (RF). The point is that it is the combination of responses defines the particular ordorant, and if you want to cover a space, having lots of overlap is good, since it allows covering a high dimensional space efficiently. In addition, local circuits mutual inhibition tends to sharpen responses to create larger differences, and receptors adapt rapidly so that smell is mostly responsive to change of smells. Borys asked how important active olfaction is. Michael answered that animals which move around towards a desired odor. The naive answer is that the animal goes up odor gradients, but in reality there are no such gradients. Instead there are eddies and vortices of air that are mixing odorants with clean air. If you have such an odor plume, then the result is a very intermittent signal as shown in the sketch. Then the animal must be sensitive to these brief pulses of odor.
Rodney Douglas then asked what is the aim of working on this? Michael answered that he hopes to make sense of these signals in the way that animals might, by taking advantage of all scales of time and space that are involved.
Sergio asked if nose shape matters? Michael answered that experiments show that even humans can do scent tracking. There was nature paper couple years ago where humans were provided a nasal prism to let them block one side or the other. That showed that binasal olfaction mattered for scent tracking. But in general not much is known about nostril shape effect on odor localizing.
Michael concluded with describing some commercially available odor sensors. There are heated sensors that respond to certain odorants, but they are slow. They are 5V suppply at 100mA, so 500mW. They have biexponential responses with quick onset of 2-3s, and offset of 15-20s for these metal oxide detectors. Another class ionizes the gas and senses UV light from this ionized gas. They are much faster (e.g. 10Hz) but very poor discrimination. There is no multi-odorant sensors with good bandwidth, low power, longevity, etc. Daniel ?? pointed out another type he used recently, that are driven with carrier frequency of e.g. 10MHz, and then the frequency shift of a few kHz is caused by odorant binding.
Next, Olivier Bertrand talked about vision, particularly in faceted insect vision. In the omatidia of such multifaceted eyes, at the bottom of each omatidia, there is a photoreceptor that continuously senses the light. There are only 700 omatidias in fruit fly eye with total of about 5000 photoreceptors covering nearly an entire hemisphere for each eye. Despite this low spatial resolution flies can clearly navigate an environment with them.
Olivier then focused on color vision in insects. Flies have something like green, blue and UV light in contrast to the red/green/blue or long, middle and short wavelength chromaphores of human eyes. You can train flies on color discrimination tasks to study their color perception.
Also, many insects (notably ants) sense polarization of light and use this for navigation.
Many insects, particularly flying insects but also beetles sense motion, and most also have optomotor responses that cause them to respond to a moving surround to stabilize it. He also presented the basic elementary motion detector proposed more than 50 years ago by Hassenstein and Reichhardt. Many of the original ideas about this correlation-based detector have held up over time, but especially there are elements like highpass filters and gain control that have been clarified by experimental work since original behavioral experiments on beetles. He pointed out that all these computations in insect brains is done with graded continuous brain signals (not spikes). He concluded with pointing out that stereo is useless for most insects (except perhaps praying mantis) but motion parallax is critical for obstacle avoidance. So insects stabilize their eyes to avoid rotation, to expose the informative motion parallax.
Tobi asked about experimental recent results. Andy Straw's lab in Freiburg has shown in VR environement. Translational flow is needed for obstacle avoidance. In butterfly there are local motion detectors. Several participants ?? are working on drone implementations of these concepts and have an implementation that avoids walls at about 1m/s (3 body lengths/second).
Next Shih-Chii Liu and Andre van Schaik discussed audition. Shih-Chii summarized the auditory pathway starting at ear drum, bones and cochlea (together the mammalian hearing organ). She then nicely summarized the biophysics of the cochlea, with its basilar membrane, inner and outer hair cells and auditory nerve. She pointed out interesting details of the timing of hte responses of these hair cells, in particular how they can follow frequencies up to about 1kHz, but above this frequency, they encode a modulated envelope of the power around this frequency.
Next Andre showed a demo of a software model of cochlea. It shows the basilar membrane displacement on the right as yellow trace. One of these locations is plotted above it. Different locations are sensitive to different frequencies.
This tool allows varying cochlea parameters, and it also visualizes inner hair cell responses, corresponding roughly to envelopes of the BM responses. The cochleagram shows energy on a logarithmic (roughly) frequency scale.
Andre pointed out that time scales are 10ms to 10s for olfaction, and up to 100Hz for vision, tactile is up to 400Hz, but audition goes to 20kHz in humans or 200kHz in bats, and some auditory nerves fire briefly up to 1kHz.
The workshop then broke for afternoon coffee break.
Next Michael Schmuker talked about olfaction. He started off pointing out that half of taste is smell, and that entire businesses such as hygiene products and airport perfume sales depend on human smell.
He pointed out the Jim Bower ;pointed out in 2013 that we now have the tools to better understand olfaction with multi-unit recordings, but we cannot record or play back smells. In chemistry, we have molecules, but they are discrete, not continuous in identity, since they consist of different hydrocarbon chain compounds. The space is high dimensional. Oderants have up to 7-8 or even 13 heavy atoms in them (C or O) and there are a huge number of permutations of these.
Receptors in nose have very broad sensitivities. A particular receptor might have 1 preferred ligand, but others overlap its receptive field (RF). The point is that it is the combination of responses defines the particular ordorant, and if you want to cover a space, having lots of overlap is good, since it allows covering a high dimensional space efficiently. In addition, local circuits mutual inhibition tends to sharpen responses to create larger differences, and receptors adapt rapidly so that smell is mostly responsive to change of smells. Borys asked how important active olfaction is. Michael answered that animals which move around towards a desired odor. The naive answer is that the animal goes up odor gradients, but in reality there are no such gradients. Instead there are eddies and vortices of air that are mixing odorants with clean air. If you have such an odor plume, then the result is a very intermittent signal as shown in the sketch. Then the animal must be sensitive to these brief pulses of odor.Rodney Douglas then asked what is the aim of working on this? Michael answered that he hopes to make sense of these signals in the way that animals might, by taking advantage of all scales of time and space that are involved.
Sergio asked if nose shape matters? Michael answered that experiments show that even humans can do scent tracking. There was nature paper couple years ago where humans were provided a nasal prism to let them block one side or the other. That showed that binasal olfaction mattered for scent tracking. But in general not much is known about nostril shape effect on odor localizing.
Michael concluded with describing some commercially available odor sensors. There are heated sensors that respond to certain odorants, but they are slow. They are 5V suppply at 100mA, so 500mW. They have biexponential responses with quick onset of 2-3s, and offset of 15-20s for these metal oxide detectors. Another class ionizes the gas and senses UV light from this ionized gas. They are much faster (e.g. 10Hz) but very poor discrimination. There is no multi-odorant sensors with good bandwidth, low power, longevity, etc. Daniel ?? pointed out another type he used recently, that are driven with carrier frequency of e.g. 10MHz, and then the frequency shift of a few kHz is caused by odorant binding.
Next, Olivier Bertrand talked about vision, particularly in faceted insect vision. In the omatidia of such multifaceted eyes, at the bottom of each omatidia, there is a photoreceptor that continuously senses the light. There are only 700 omatidias in fruit fly eye with total of about 5000 photoreceptors covering nearly an entire hemisphere for each eye. Despite this low spatial resolution flies can clearly navigate an environment with them.Olivier then focused on color vision in insects. Flies have something like green, blue and UV light in contrast to the red/green/blue or long, middle and short wavelength chromaphores of human eyes. You can train flies on color discrimination tasks to study their color perception.
Also, many insects (notably ants) sense polarization of light and use this for navigation.
Many insects, particularly flying insects but also beetles sense motion, and most also have optomotor responses that cause them to respond to a moving surround to stabilize it. He also presented the basic elementary motion detector proposed more than 50 years ago by Hassenstein and Reichhardt. Many of the original ideas about this correlation-based detector have held up over time, but especially there are elements like highpass filters and gain control that have been clarified by experimental work since original behavioral experiments on beetles. He pointed out that all these computations in insect brains is done with graded continuous brain signals (not spikes). He concluded with pointing out that stereo is useless for most insects (except perhaps praying mantis) but motion parallax is critical for obstacle avoidance. So insects stabilize their eyes to avoid rotation, to expose the informative motion parallax.
Tobi asked about experimental recent results. Andy Straw's lab in Freiburg has shown in VR environement. Translational flow is needed for obstacle avoidance. In butterfly there are local motion detectors. Several participants ?? are working on drone implementations of these concepts and have an implementation that avoids walls at about 1m/s (3 body lengths/second).
Next Shih-Chii Liu and Andre van Schaik discussed audition. Shih-Chii summarized the auditory pathway starting at ear drum, bones and cochlea (together the mammalian hearing organ). She then nicely summarized the biophysics of the cochlea, with its basilar membrane, inner and outer hair cells and auditory nerve. She pointed out interesting details of the timing of hte responses of these hair cells, in particular how they can follow frequencies up to about 1kHz, but above this frequency, they encode a modulated envelope of the power around this frequency.Next Andre showed a demo of a software model of cochlea. It shows the basilar membrane displacement on the right as yellow trace. One of these locations is plotted above it. Different locations are sensitive to different frequencies.
Andre pointed out that time scales are 10ms to 10s for olfaction, and up to 100Hz for vision, tactile is up to 400Hz, but audition goes to 20kHz in humans or 200kHz in bats, and some auditory nerves fire briefly up to 1kHz.
The workshop then broke for afternoon coffee break.
Comments
Post a Comment