In brief, we consider the channel as frequency-selective channel if
the frequency of the signal is larger than then frequency of channel
No offense, but that ought to win the prize for the least-accurate definition of frequency-selectivity I heard ;-)
What you mean is channel is frequency-selective if the bandwidth of the signal is larger than the coherence bandwidth of the channel.
So, does using the spreading spectrum change the frequency-selective channel into frequency-flat channel? is that right?
No (and yes). As the name says, spreading the spectrum makes the occupied bandwidth of your signal larger. So ending up in a situation that the signal undergoes frequency selective fading increases a lot. The point is not the spreading, the point is that you spread the same information over a wider bandwidth and then at your receiver, you can coherently combine all these contributions. Therefore, after your combiner (which is probably what you meant anyways), the channel you'll experience will have a significantly lower probability of fading, since it combines the channel contributions over a wide range of frequencies with fade independently (frequency-selective) and thus are very unlikely to fade all at the same time.
- After spreading, your signal undergoes frequency-selective fading.
- After despreading/combining, your signal experiences a channel that is much more stable/closer to a line of sight connection.
- Since your frequency content has been "aggregated" you don't need to worry about frequency-selectivity after your combiner anymore (assuming all went well, e.g., you are perfectly synchronized).