Thank you all for the information.
I did some noise tests with my DS1054z, in all tests the probe head are floating in air and aren't directly connected to any signal source. Here are the vpp peak-peak value:
Disconnect probe: 10.4-11.2mv (this is the oscilloscope internal noise, normal?)
Connect probe with GND spring, probe head on antistatic grounded mat: 10.4-12.0mv
Connect probe with GND spring, probe head on antistatic UN-grounded mat: 17-18mv
Connect probe with GND spring, probe head not on antistatic grounded mat: 12-30mv depend on the different position
Connect probe without GND spring, probe head on antistatic grounded mat: 12.8-14.4mv
Connect probe without GND spring, probe head on antistatic UN-grounded mat: 23.8-24.8mv
Connect probe without GND spring, probe head not on antistatic grounded mat: 13.6-56mv depend on the different position
Connect probe with std hook cap, probe head on antistatic grounded mat: 46-50mv
Connect probe with std hook cap, probe head on antistatic UN-grounded mat: 106-110mv
Connect probe with std hook cap, probe head not on antistatic grounded mat: 34-200mv depend on the different position
So, when working on the grounded anti-static mat and the probe with ground spring, I got the almost noise free signal input. In conclusion:
1. Grounded antistatic mat significantly reduce the noise level. I'm planning buy a larger antistatic mat that can cover my whole table
2. GND spring can help reducing the noise level and get almost perfect signal, but not always convenient. I think in most cases, the bare head without GND spring should be good enough.
3. Try stay away from the std hook cap, which significantly amplified the noise level up to 200-300%. I think this is probably because the hook cap doesn't have a grounded metal wrap around the probe needle, thus make the unwrapped needle much longer than original. This "extended antenna" is more sensitive to the common noise waves around us, and receive more energy from those noise. I wonder if adding a grounded metal wrapper could improve the performance of the hook cap. Will do some test later.