Microphone Mechanomyography Sensors for Movement Analysis and Identification

• Published in: 2022 International Conference on Advanced Robotics and Mechatronics (ICARM) pp. 118 - 125
• Main Authors: Paszkiewicz, Filip P., Wilson, Samuel, Oddsson, Magnus, McGregor, Alison H., Alexandersson, Asgeir, Huo, Weiguang, Vaidyanathan, Ravi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 09.07.2022
• Subjects: Legged locomotion; Muscles; Optical fibers; Optical variables measurement; Robot sensing systems; Vibration measurement; Vibrations
• DOI: 10.1109/ICARM54641.2022.9959672

Muscle Activity is one of the most important signals in analysis of human movement. Commonly, optical, or inertial measurement systems are paired with electromyography (EMG) electrodes to provide an information suite for analysing human movement. In many applications in uncontrolled conditions, EMG signal acquisition may be challenging due to perspiration, changing skin impedance or the need for a reference signal. A complementary muscle monitoring modality that has been shown to overcome some of the limitations of EMG is mechanomyography (MMG), which measures the vibrations of muscle fibres during contraction. In this work, we show that MMG combined with data from an inertial measurement unit (IMU) in a novel sensing suite can be used in monitoring muscle activity during common gait activities. Data were collected from a cohort of 9 volunteers. MMG-IMU movement profiles are presented for cyclic movements e.g., walking, or ascending stairs and noncyclic movements e.g., standing up or sitting down. A multi day study was conducted which demonstrates that data collected over several days can be used to generate a general movement profile. Average correlation for leave-one-out analysis between 4 days and a 5th day was found to be 90% for sitting down motion and 64% for standing up motion. Lastly, MMGIMU sensor fusion was shown to be well suited for classification of daily movement using Support Vector Machines. With the addition of MMG muscle data increasing classification accuracy by 3%, from 91% for IMU to 94% for MMG-IMU.