Abstract: In this paper,a current-mode linearly classifier which can classify linearly non-separable data is proposed.The proposed classifier circuit is composed of trapezoidal activation function circuits and linearly weighted circuits.Linearly weighted circuit is achieved by fully balanced differential transconductor and trapezoidal activation function circuit is mainly comprised of threshold circuits.In order to achieve the classification of the linearly non-separable dataset,the weighting coefficients can be obtained based on Fishers linear discriminant analysis by MATLAB software.The proposed circuit is simulated and analyzed by PSPICE.The results show that the proposed circuit recognizes the characteristics of simple structures with high accuracy and low power consumption.The proposed classifier circuit has potential applications in various fields such as pattern recognition,neural networks and artificial intelligence.