Fusion of sensory and mechanical testing data to define measures of snack food texture

The multivariate nature of the texture quality space of a tortilla corn chip product is investigated by fusing textural information from three diverse sources (sensory panel analysis, visual appraisal and mechanical breaking force) via a principal component analysis (PCA) model. In preparation for the PCA model, a combined wavelet transform and histogram analysis method is developed to extract texture-related features from the mechanical breaking force instrument. A physical interpretation is provided for each of the two significant principal components of the PCA model, which provides insight into the nature of the full texture quality space.There are several practical implications of having obtained these two robust textural features. Firstly, they can be used to better define multivariate specification regions. Secondly, it is shown that these two features can also be extracted via PCA using only the visual appraisal and mechanical breaking force data, thus enabling at-line statistical process control based on these two fused variables. Thirdly, these two features provide robust quality variables for developing on-line inferential sensors from process measurements.