Since microspeakers are widely used in mobile phones, high power is usually applied to obtain sufficient output sound pressure. However, the electric power is almost converted into heat, leading to the thermal problem in microspeakers. Compared with large loudspeaker, microspeakers are smaller and the under yoke is relatively closed, causing the heat transfer more complex. In this study, a three-stage nonlinear thermal model was proposed for analyzing the thermal behavior in microspeakers. The inside air is a buffer area between the voice coil and magnet, and modeled as a middle stage of the heat transfer. The forced convection is still significant in microspeakers while the eddy current can be ignored. In order to obtain the thermal parameters of the model, a corresponding parameter identification method was put forward. The basic linear parameters and forced convection parameters were all obtained by measuring and fitting the temperature curves of voice coil at different single tones. A series of experiments were conducted to verify the proposed model and parameter identification method, and the results showed good aggrement between the measured and predicted temperature curves for different input signals. The proposed model was valid and accurate, and may be helpful for the design and application of microspeakers.