Electrical Impedance Myography Measurements for Gesture Recognition Data Normalization

Electrical Impedance Myography (EIM) is a non-invasive method based on applying a low-intensity alternating current to muscle tissue and measuring the resulting voltage over these muscles in various frequencies, EIM has been recently used to recognize human hand gestures by measuring resistance (R) and reactance (X) changes during muscle contractions and relaxation. In hand gesture recognition, the level of muscle contraction and muscle shape typically varies within and between subjects for the same gesture; additionally, the subjects’ hands have different sizes and muscle shapes; normalization techniques are used to compensate for these differences and find a new specified data range from the existing data, this enhances the accuracy of the results and avoids the size effect. In this work, four normalization methods: Z-score, Min-Max, Decimal Scaling and Median & Median Absolute Deviation methods were applied to the multi-frequency EIM data measurement for 9 American sign language gestures. The result shows that the z-score is effective for normalizing EIM data with 100% training accuracy and 94.5% testing accuracy using an extreme learning machine classifier, the z-score method preserved a single value per muscle per patient; it helped to normalize across types of muscle and eliminated a possible confounding in the analysis variable. The results show that the z-score normalization is effective for hand gesture recognition using EIM data.