A memlumor is an innovative neuromorphic luminescent device with a state-dependent photoluminescence quantum yield (PLQY) designed for optical neuromorphic computing applications. Metal halide perovskite memlumors leverage charge trapping and photodoping to modulate the PLQY, making it dependent on the excitation light history. Here, we demonstrate the ability of perovskite memlumors to classify time-dependent binary optical signals on sub-microsecond timescales. Sequences of laser pulses (representing ones) and gaps (absence of pulses representing zeros) were used to excite photoluminescence in MAPbBr₃ and triple cation perovskite films. By reading only the time-integrated PL signal, we completely recognized the time-dependent input patterns containing 5 bits of information. The potential of perovskites for applications in optical reservoir computing based on their complex and diverse photophysics is discussed.