Spatial audio reproduction is essential to create a natural listening experience for digital media. Majority of the legacy audio contents are in channel-based format, which is very particular on the desired playback system. Considering the diversity of today’s playback systems, the quality of reproduced sound scenes degrades significantly when mismatches between the audio content and the playback system occur. Primary ambient extraction (PAE) is an emerging technique that can be employed to solve this pressing issue and achieve an efficient, flexible, and immersive spatial audio reproduction. This thesis investigates the performance of existing PAE approaches and develops new techniques that not only improve the performance, but also enhance the robustness of PAE approaches in dealing with more complex signals encountered in practice. Objective and subjective evaluations validate the feasibility of applying PAE in spatial audio reproduction systems (as also highlighted in the headphone system).