This thesis investigates the performance of software defined radio in reconstruction of a coded speech signal in presence of channel errors, taking into account end to end communication. At the transmitter, the recorded author’s voice is encoded using linear predictive coding algorithm, where speech parameters such as linear predicted coefficients, pitch, voicing and gain parameters, are extracted from the speech signal. These parameters are sent to linear predictive decoder to model the speech signal from its parameters. The output from source encoder is sent to channel encoder such that, the digital encoded speech data is protected using linear block codes algorithm to provide error protection to bit stream before transmission to the communication channel. Receiver’s blocks or algorithms to curb multipath interference, intersymbol interference, timing offset and carrier offset are based on adaptation. Steepest descent adaptive algorithm is used to design the entire algorithms to run the software receiver. Therefore, steepest descent algorithm is implemented in down conversion, carrier recovery, clock recovery, equalization and correlation. All algorithms running the software receiver are theoretically discussed and implemented in MATLAB software. The results obtained after simulating the whole receiver block in terms of symbol error rate, mean square error and bit error rate are recorded and analysed to investigate how channel errors affect software receiver while reconstructing a coded speech signal.