September 3, 2017
Surface TensionToday in fluid mechanics class we learned about surface tension and it's some pretty crazy stuff. It does super weird things and it makes me wonder what the world would be like without it. A couple classes ago my professor was talking about how, since we're carbon based life-forms, H2O is our "universal solvent". He continued to mention that if we were, for example, silicon based, perhaps we would have a different universal solvent. Would it still need to have similar surface tension properties?
Sound RecordingMy friend was recording a lecture on his phone and he found that his professor's voice was really faint but all the rustling of clothing was really loud. I instinctively thought maybe a filter ccould help solve the problem, but then I realized that probably wouldn't help that much and that a directional mic was probably the most effective solution.
Analog Fourier TransformerDeciphering speech made me wonder how our brains can so effectively analyze sound. I remembered from biology and a Music and the Brain class I took that the ear essentially has a built-in Fourier/frequency analyzer via the design of the cochlea. Could this be replicated in either mechanical sensors or electronic hardware to lessen the load on the software to recognize speech? Perhaps this wouldn't save all that much computational time given the relatively light demands of FFT, but maybe other filters could be utilized to identify explitives and whatever other consonant sounds exist.
Then again, I think most of the difficulty in speech recognition is putting the words "options" together in a way that makes sense. For example, the sentence "Dogs and cats are good pets" might be misheard as "Dogs and bats are good pets", but we can interpret it correctly based on context. I bet this is where the real interesting computational stuff comes into play and I doubt any amount of hardware could do that!