Due to the limitation of the radio frequency (RF) spectrum, it is increasingly more difficult to support billions of wireless devices in the age of Internet-of-Things. Consequently, many recent wireless indoor communication systems have been developed using free space optical (FSO) communication technologies that exploit the extremely large light spectrum to transmit data. This dissertation contributes three physical layer techniques to an FSO system called WiFO, which is integrated with the existing WiFi transmission. (1) We proposed a Pulse Amplitude Modulation (PAM) decoding scheme that estimates the channel parameters and determine the optimal decoding thresholds to minimize the average bit error rate (BER). The conditions on channel parameters for which signal recovery is impossible, are also determined. (2) A pre-shaping technique called Adaptive Sending Duration Algorithm (ASDA), that modifies the durations of input bits dynamically, is proposed to be used at an LED-transmitter. Simulation results show that ASDA working with equalization successfully improves the performance of BERs. (3) We propose a decoding scheme called Memory Decoding Algorithm (MDA) used at a receiver and show that it effectively reduces the bit error rates via maximum likelihood decoding principle.