Well, I don't know how to confirm to myself that the words which I'm writing are my own words. Now, maybe those words are based on what I have read recently because if not there won't be anything to say. Should I for example just comment in what other researchers say which I'm paraphrasing what they wrote or can I also write as if I was talking (words which are based on what I have read)?
I don't know what the purpose of your writing is, so the answer really depends on that. It sounds like you're writing some kind of review of prior work (either as a survey or as part of a paper) ? In that case, the goal here is not to regurgitate (in your words or via paraphrase) what others say. Rather, you should be reading what they say and thinking about it (and seeing if you're convinced by it).
Only after that should you even attempt to describe the work. And when you do so, put all reference material away. If you can't describe someone else's work without referring to it, then you don't really understand it yet. In this way you'll ensure that you use your own ideas/thoughts to express what's gone before.
I second @Suresh's remarks, but/and the situation is truly more complicated, as your explanation of your question correctly indicates.
For example, complicated, sophisticated things are often so easy to mis-represent that it seems as though there's almost a unique path, a unique narrative, that is correct. And it may be non-trivial to learn what this path is. Thus, in effect, the phrasing of a high authority is not only "burned into one's retinas", but, also, seems the only correct thing to say.
So, first, do not try to "say something else", just for the sake of avoiding "quoting", because that seeming paraphrase may be wrong, or deficient, or ...
But, at the same time, extended passages should not be quoted, unless put in quotation marks. It's best to internalize ideas well enough to give "the standard definition" and such things, even if that ends up sounding very similar to other sources. Similarity is inevitable in many cases.
As in many scenarios, honesty is a clarifying criterion. That is, is what you write coming out of your own head, even if it resembles other sources, or are you quite literally copying? The latter is not so good.