I've been an Amateur Radio operator for 40 years (WA7EMS is my call sign, a "Vanity call" and yes, the phonetics DO mean what they say... )
What you are intending to do is difficult because of the Average "noise" level where noise is defined as all detectable energy. Providing selectivity at 900 MHz (Approx) can be done by several different methods, the simplest is the antenna and a 1/4 wave antenna is usually the choice for several different reasons that any RF engineer can point out. There are several problems with that issue too. The principal one is the selectivity of a 1/4 wave antenna isn't the greatest. A "Hotcarrier" diode... HP 5082-28XX series diodes or an OLD 1N23 microwave diode will make a great detector... for anything from 400 Mhz and up, the "sorta' limit is the anrtenna length as it is the only selectivity in circuit. Convert the RF to DC with the detector and apply some DC gain with an op-amp... You could use an LM358 here but I am real partial to R-R I/O Op-Amps and I would use an LMC662. It is a single or dual power supply type of op-amp that can swimg from Vcc to ground in single ended types of circuits and can operate to ground on the inputs whare some types of bo-polar types will latch up. So now you have an amplified broad band noise detector with a small antenns (1/4 wave @ 900 MHz is about 3 inches) and even that is do-able but to differentiate between SMS and phone conversations is nearly impossible and totally so when you consider the other sources of EMI (a simple message might well be 100 mS for the phone to receive the information and respond to the cell site for the message before announcing the call to you as the data rates are rather high < 100K baud. IMO you will have to be real clever to accomplish your stated goal.
Bob