Background: Rapid and easy technology which can mimic the tongue for the simultaneous
perception of several tastes based on sensory analysis and mathematical statistics is sorely needed.
Methods: Joint voltammetry technology was developed to qualitatively and quantitatively analyze
four basic tastes namely sweetness, saltiness, sourness and bitterness with the multi-electrode array.
Four taste stimuli were corresponded to four tastes. Cyclic Voltammetry (CV), Differential Pulse
Voltammetry (DPV) and Square Wave Voltammetry (SWV) were employed. The original voltammetric
signals were transformed by Continuous Wavelet Transform (CWT) in order to reveal more feature
information for sensing taste stimuli. Joint voltammetry was applied via the combination of voltammetry.
The data of feature points from the transformed signal as the input were used for neural network model.
Results: Layer-Recurrent neural network (LRNN) could effectively identify the types of stimuli. The
accuracies of the training set and test set by joint voltammetry were both higher than that of regular
voltammetry, confirming that Back Propagation neural network (BPNN) could quantitatively predict
single taste stimulus of the mixture.
Conclusion: Joint voltammetry technology had a strong ability to sense basic tastes as human tongue.