The correlation and value of multimodal lie features are researched. Due to the lack of single-modal features, the accuracy of lie recognition will be affected. Therefore, combined with objective speech and physiological signals, the lie correlation of formant frequency, speech energy, Mel-frequency cepstrum coefficients, Gammatone frequency cepstrum coefficient, heart rate and other multimodal features based on statistical principles is analyzed. Experiments on ten subjects showed that the second and the third formants, Gammatone frequency cepstrum coefficient and heart rate have higher lie correlation, which provides theoretical and data support for the selection of lie features.