VoIP calls are sensitive to several impairments, such as delay and packet loss. One way to overcome these problems is by adaptively adjusting application-layer parameters to keep a minimum speech quality level. At the heart of self-adaptive systems lies a feedback loop, which consists of four key activities: monitoring, analysis, planning, and execution. Nevertheless, the existing adaptive approaches to QoS control of VoIP do not explicitly exhibit this feedback loop. Bringing it to surface can help developers in designing more robust and human-independent VoIP systems. This survey presents a comprehensive review of the current state-of-the-art research on speech quality adaptation of VoIP systems at the application layer and some research challenges on this subject.