Research data supporting [Ultrasensitive Textile Strain Sensors Redefine Wearable Silent Speech Interfaces with High Machine Learning Efficiency]

This work encompasses five related datasets, accessible via an open-source link provided at the end of the manuscript: 1. Dataset1_20 Frequently Used Words: This dataset contains signals of the 20 most frequently used words (10 nouns and 10 verbs) collected from participants, with 100 samples per class. Each sample of a word is represented in a row, with the last number in each row indicating the class label for that word (the same applies to the following datasets). 2. Dataset2_Confusing Words: This dataset includes 5 pairs of 10 words with similar pronunciations that are easily confused, with 100 samples per class. 3. Dataset3_Different Reading Speeds: This dataset comprises signals of 5 long words read at three different speeds: fast, medium, and slow, with approximately 33 samples for each word at each reading speed. 4. New User Generalization Test: This dataset contains signals of 5 commonly used words (included in Dataset1) collected from three new users, with 50 samples per class. 5. Noise Injection Data: This dataset includes around five minutes of silent noise signals (containing physiological noises such as breathing and swallowing) recorded in the absence of speech.