Collecting environmental sound by utilizing high-end mobile phones provides us opportunities to capture rich contextual information in real world. The gathered information can be used for various purposes, ranging from academic research to livelihood support. Furthermore, mobility of mobile phones opens a door for easily forming a dynamic sensing infrastructure, in order to gather fine-grained, but still large-scale data from both spatial and temporal perspectives. However, collecting, analyzing, storing, and sharing of sound data usually involve large energy consumption than scalar data, and like any battery-operated device, mobile phones also face the reality of energy constraints. Because people's first priorities are naturally to use mobile phones for their own purposes, there are occasions when people will not be inclined to allow their mobile phones to be used as sensing devices fearing that they will run out of batteries. Therefore, our research focuses on energy-efficient sensing, to reduce average energy consumption and to extend overall system lifetime. In this paper, we propose a node scheduling scheme for mobile nodes. By applying this scheme, optimized sensing schedules (ACTIVE/SLEEP duty cycles) will be periodically generated at each node. Following the provided schedule during sensing, energy-efficiency can be realized while original Quality of Service (i.e. coverage rate) is retained. Unlike most previous works which were based on ideal binary disk coverage model, our proposal is designed under a probabilistic disk coverage model which takes the characteristic of sound propagation into consideration. Furthermore, this is the first scheme that is adaptable to large-scale mobile sensor networks where topology dynamically changes. An accurate energy consumption model is adopted for evaluating the proposed scheme. Simulation results show that our scheme can reduce up to 48% energy consumption in an ideal environment and up to 31% energy consumption in a realistic environment. The robustness of our scheme is also verified against different type of sensing terrains and communication environments.