Accurate identification of the working conditions of froth flotation remains challenging because of the inherent chaotic nature of the underlying microscopic phenomenon. The froth surface is generally used as an effective indicator of the working condition and performance of flotation. In this study, we developed a novel method for determining the complex working conditions of flotation through statistical modeling of froth images. Gabor wavelet transformation was used for modeling because of the Optimal localization properties in both spatial and frequency domain of the Gabor functions. The characteristic parameters of the probability density functions of the Gabor filter responses of the froth image, rather than conventional statistics (mean and variance), were then modeled using the empirical probability distribution models, t location-scale and gamma distributions. A simple learning vector quantization-neural network (LVQ-NN) was adopted to obtain an effective classifier for identifying the working conditions of froth phases under different production phenomena. The proposed model was validated through experiments on a bauxite flotation plant located in China and compared with commonly used determination methods. (C) 2015 Elsevier Ltd. All rights reserved.