A phoneme is the smallest meaningful unit of sound in a language. A meaningful sound is one that will change one word into another word. For example, the words cat and fat are two different words, but there is only one sound that is different between the two words - the first sound. That means that the “k” sound in cat and the “f” sound in fat are two different morphemes.
Now consider the words skin and kin. Both words have an “k” sound, but they are really slightly different sounds. The “k” sound in skin is softer that the “k” sound in kin. Those two sounds are not phonemes in English. They are what are called “allophones,” which are just variations of a phoneme. In another language, however, those two sounds could be phonemes.
That means that two groups of sounds which have only those two “k” sounds as a difference between them would be two different words. Pretending that the two groups of sounds were “kin” (with a hard “k”) and “kin” (with a soft “k”), you'd have two different words with two different meanings. You can try to say those two words, but you'll probably end up saying kin, but starting the other word with a “g” sound as in gun. The reason is that we have a hard time hearing and repeating sounds that are not phonemes in our language.
Shortly after birth, a baby begins to learn the phonemes of the language used around him. It's part of what he learns as he learns language. We don't have to teach babies those sounds; they are simply programmed to learn them as they interact with people. (It's one of the reasons it's good to talk to babies a lot.) As children continue to learn language, they aren't consciously aware that words they are learning are made up of separate and very distinct sounds.