I am not sure if I am asking a question in the right site. I don't know know if I am even asking a biological, linguistic or physics question. But I recently started learning about language and its structure and I just got struck by a sudden question that seems so basic yet so difficult to exactly answer. This question is that what exactly is the difference between a sound that is made by the human voice box when forming words in comparison to sound made by non-living object, for instance a sound made by hammer hitting on metal object, bouncing ball? Both are essentially sounds, but when makes on a phone/phoneme and the other just plain sound. Also call all sounds made by the voice box be called a phone/phoneme such as a sharp cry by a baby, a shriek such as "ahhhhhh" made by someone when scared? If not, then what are the criterias or demarcating lines that separates sound a phone/phoneme from just plain sound made by any object in the universe? Thanks.
Forget about phoneme, until you get "phone" versus "sound". A "phone" is a specific kind of sound: it is an interpretation of sound as part of a linguistic system. Wind and bird noises are just sounds. There are many sounds that cannot possibly be produced by human anatony. There are also many sounds produced by the human vocal apparatus which are not "phones", for example yawning, or "the raspberry". Some noises are physically indistinguishable from language phones, like "Hmmmmm" contains a very long version of the phone at the end of the word "spasm". The difference, other than the length of the sound, is how the listener interprets the sound. We understand that "Hmmmm" is not part of language, so the m-like thing is not a phone (it's not language sound). We might interpret the sound [ʔai] as a 1st person pronoun "I", when said by an English speaker, but it might also be a non-linguistic interjection analogous to "uh" (you can quickly figure that out from context – that speaker actually pronounced the pronoun different from the interjection).
The sound does not have to be produced by a human: it can be synthetic, or produced by parrots. I think (at least given intelligible parrot speech) that parrot's produce sounds which we interpret as phones (because phone status is about how listeners interpret the sound).
A phone is a maximally accurate transduction of sound to linguistic sound categories: it is (ideally) language-independent. It is not interpreted in terms of systemic notions like "contrast", "minimal pair" and so on. "Phoneme" is a way that some linguists analyze rule-governed distribution of phones in a language, for example aspirated vs. non-aspirated stops in English are different kinds of phones (p vs pʰ), but there is a rule that tells you when you get one vs. the other, so it is said that there is a single phoneme, usually /p/, and a rule that derives the other variant [pʰ] in a particular context (stress-foot initial position).
The comment by OP raises an interesting untested question. Can we introducing novel segments that are humanly impossible to produce, but which are otherwise just like regular language sounds? For example, replace all instances of "z" with a perceptibly different sound: can a person parse [sɪp], [⁂ɪp], [kæts], [dɔg⁂], [bʊʃɨ⁂]? The Substance-Free theory is that this poses no problem, the Substance-Dependent theory predicts that the system can't be perceived as language (if [⁂] is sufficiently bizarre).
There is already a substantial body of evidence that the set of language sounds is enormous, where the sound properties of [t], [tʰ], [d] vary greatly across languages. If there is a small, pre-defined set of phones with exact values available for human language, the prediction is that [tʰ] would sound the same in every language that has [tʰ], but it is well established that this is not the case.