Backscatter communication is essential in applications where wireless and batteryless sensors are demanded. Backscatter communication system may suffer short communication range because its sensors transmit signals by means of the reflection of incident radio wave that is provided by counterpart interrogator. The use of multiple interrogators, forming multiple interrogation zones, is the usual practice to extend the range of backscatter communication system. However, if a sensor receives more than one radio waves-one from its counterpart interrogator and others from neighbor interrogators-the sensor produces modulated backscatters to all the interrogators, causing interference to neighbor interrogators. This inter-zone interference is particularly problematic when multiple sensors send stream data concurrently. This paper proposes a real-time separation of collided backscatter signals based on their statistical independence. The proposal uses a priori probability of backscatter signal to extract the independent carrier phase angles and the independent amplitudes from IQ signal observation. The measurement of channel state information with predefined symbol pattern in packet is not needed. The proposal is evaluated both in numerical simulations and a physical experiment using prototype backscatter sensors and a software-defined interrogator. The results show that collided backscatter signals can be separated in real-time without degrading the packet error rate of each signal.