I got the package sphinx-2 from the repositories using my newest scripts which I am enjoying. I have some trouble remembering names that don't completely describe what I want. So I have extended my scripts to replace apt-get install with something I can easily remember.
It seems that the idea of Fourier analysis may not be the best tool to understand speech. I know that this is the common way to look at it , but I think that my n-D software should make short work of the first level, which is to assign sounds to a specific set of sound-a-likes (homophones). After that it seems that the concept of the inverse of English language programming would be applicable.
I haven't tried this before so it should provide some entertainment and a good puzzle to solve.
The association of sounds to homophones seems that it should resolve easily and I wonder why it is considered so tough and there is no open method available. It also seems that there should be a list of homophones somewhere so I don't have to duplicate that research. It would seem that I could search a dictionary that had phonetics to find the matches if a list doesn't exist. I think the reference below which I downloaded has everything I need for homophones included.
From The Collaborative International Dictionary of English v.0.48 [gcide]:
Not \Not\, adv. [OE. not, noht, nought, naught, the same word as E. naught. See {Naught}.]
I have a plan which involves a strange perspective. It seems to me that the sequence of muscle contractions should have a direct one to one correspondence between some aspect of the sound and the corresponding muscle sequence. It seems that the frequencies are merely incidental to the communication. It seems another case where the whole is assumed to be a sum of recognizable and separable mathematical elements. It might have been a real coop to have a correlation of Fourier analysis, but it seems that like all other problems they contains an infinite set of possibilities. Many of these methods that try to resolve problems with a mathematical system seem to fail miserably. It also seems that I add many more levels of complexity to this it will resolve easier. Things like lip reading, and music. It seems that, as I have experienced before, a matrix like solution of a set of variables requires a sufficiently large unique set of equations of the variables to equal the complexity of the problem or simply a set of unique equation relationship that equals the number of variables. In this case it would seem that the number of muscles involved would be the number of variables that need to be considered.
ADDED: After some thought this seems to be a good approach. The medium acts as a transport layer in the same way as Asynchronous Transfer Mode (ATM) works with the internet. It makes no difference how the bits are moved from A to B, and in this case from brain to brain. If I think cow and say cow then they hear cow and think cow, It is { idea ⇒ speech ⇒ waves ⇒ ear ⇒ nerves ⇒ idea} and so even the muscle contractions and relaxations could be considered a transport layer. Also the same effect is achieved in this way: { idea ⇒ write ⇒ words ⇒ blog ⇒ eye ⇒ nerves ⇒ idea}.An idea is a singular selection from a set of possible states that could exist. As always when dealing with infinities it is the significant elements that have meaning and so in this case I could just as well assign an Item Unique IDentification (IUID) to the concept itself on the originating and receiving end. The problem may arise that there is no IUID on the receiving end and in this case it would be nonsense to the receiver. Okay, so some of what I say is nonsense at the origin which is a separate issue.
The ability to mimic is inherent in the hardware. In the case of a person, if the hardware is the same ( generally), then if I simulate speech and it matches in my mind, it should match in the receiver. This is not always the case however. If the person mimicking has a tonal loss then they will recognize and repeat a wrong model. Training and application of specific phonemes arise in culture and it is common that certain vocal patterns are not trained at birth by imitation and as a result become nearly impossible to train at a later age.
Communication of information and concepts between people need not proceed by the intervening medium of stick figures or sound waves or even visual cues. It seems that if sound is ambiguous to the person involved, a universal qualifier would serve as well.
This seems to lead to the concept of a universal language. This seems reasonable. A reference is all that is required. In the case of my blogging I have started developing a language interaction which allows me to streamline my common actions by simply typing something like wiki?ATM and as a result it replaces it with the appropriate reference to Wikipedia by finding the reference, creating a link, and substituting the commands to achieve that. I can use my English language programming to achieve that via a script.
ADDED MORE: It seems to me, and it may be wrong, but sound is actually 3 dimensional and that it is stored as a one dimensional or two dimensional format. There has been something that was on the edge of my understanding and this may resolve something that was used by accident and very effective in its application, however I am quite certain that those who applied it were completely unaware that they had stumbled upon a fact about sound that was present and not understood. I will do some analysis to determine if I am correct. When doing 3D two camera are used to have the same perspective as the eye and a stereo microphone also incorporate some of the 3D of sound. The problem is that though it does simulate the state of measurement , it is removed from interaction and as such loses an entire infinity in representation in that manner.
Since speech or communication originates from a singular event expands to the muscle actions then becomes a broadcast effect in 3D and is collected and transformed and finally identified seems to imply that the actual communication is not a complex event at all and that the method of transfer is the complex element and it is completely arbitrary in its implementation and measure.
So the communication of minds with words is closer to the concept framework, however it suffers from the same arbitrary establishment of stick figure medium of exchange. It seems that somewhere under all the complexity is a different complexity which mirrors this process and determines the structure of the communication. It would seem that the [vector in many dimensions] that is expression extrapolates through a point and emerges in the structure of the mind. Analysis continues, no firm testable conclusion yet.
ADDED EVEN MORE: So it seems that I have developed an understanding of this. Firstly the representation of a sound as frequency distribution is not a description of the sound, it is a mechanical method to duplicate the effect. This is vastly different than the understanding of what it represents. This tells me that mathematical analysis is a ridiculous method of dealing with any subject. I can duplicate the nature of a waveform as a sequence of sound level pressures in time. This replicates the signal, but does it describe its meaning? Absolutely not. It is just another transport layer. I can draw a stick figure by turtle program turn left, pen down, move, ....pen up. This does not describe the complexity within. The objects are continuously factorial and even in a communication as small as 70 unique relationships expands in potential complexity to 10^100. It is that position in n-Dimensional space that is unique. Within human comprehension I am guessing only that this n-Space is limited at some where in the area of 10,000 dimensions in context and a clarity between infinity and zero of 10,000 elements. This then resolves to 100,000,000 concepts within a possible context state that is perhaps only 100 times more complex than that. This is certainly a massive number since it is a differential factorial array.
So then I can define the concept as a value which represents a vector in n-Space. It would be a 16 bit integer array of anywhere between 1 and 10,000 dimension from a selected group of dimensions and thus I can , I suspect define a universal ID to the common concepts and project that into mind space to simulate the interaction. The complexity is probably 10^10 factorial, but that is not infinite, though it is a very large number compared to even the atoms in the entire Universe.
It makes me think of a science fiction concept that crossed my mind today, "The Body Left Behind" by a being that became detached from its existence by discovering how to become immortal in a new form and what happens is that the body is left behind with the physical mind and the memory of the other, but stuck in physical form to live out the life of the organism and die as the "body left behind". A bit freaky even for my imagination. It seems in science fiction the concept is that the "spirit or being" leaves the body and it falls into a heap, but that seems a bit hoky as even if the mind was gone, ( which some have commented that mine is already MIA ) that would not take the impulse that drives the heart, unless the mind controls the heart, which it certainly does not. That is about as far as I will speculate on that particular existential fantasy excursion.
0 comments:
Post a Comment