IEEE SENSORS JOURNAL, VOL. 13, NO. 7, JULY 2013 2499 Mechanomyography Sensor Development, Related Signal Processing, and Applications: A Systematic Review Md. Anamul Islam, Kenneth Sundaraj, Member, IEEE, R. Badlishah Ahmad, Nizam Uddin Ahamed, and Md. Asraf Ali Abstract— Mechanomyography (MMG) is extensively used in the research of sensor development, signal processing, characteri- zation of muscle activity, development of prosthesis and/or switch control, diagnosis of neuromuscular disorders, and as a medical rehabilitation tool. Despite much existing MMG research, there has been no systematic review of these. This paper aims to deter- mine the current status of MMG in sensor development, related signal processing, and applications. Six electronic databases were extensively searched for potentially eligible studies published between 2003 and 2012. From a total of 175 citations, 119 were selected for full-text evaluation and 86 potential studies were identified for further analysis. This systematic review initially reveals that the development of accelerometers for MMG is still in the initial stage. Another important finding of this paper is that sensor placement location on muscles may influence the MMG signal. In addition, we observe that the majority of research processes MMG signals using wavelet transform. Time/frequency domain analysis of MMG signals provides useful information to examine muscle. In addition, we find that MMG may be applied to diagnose muscle conditions, to control prosthesis and/or switch devices, to assess muscle activities during exercises, to study motor unit activity, and to identify the type of muscle fiber. Finally, we find that the majority of the studies use accelerometers as sensors for MMG measurements. We also observe that currently MMG-based rehabilitation is still in a nascent stage. In conclusion, we recommend further improvements of MMG in the areas of sensor development, particularly on accelerometers, and signal processing aspects, as well as increasing future applications of the technique in prosthesis and/or switch control, clinical practices, and rehabilitation. Index Terms— Mechanomyography, muscle characteristics assessment, prosthesis control, sensor development, signal processing. I. I NTRODUCTION M USCLES make up a large part of the body and are subject to many injuries due to accident, excessive use, inflammation, diseases or infections, and suffer from side effects of certain medications. They can be affected by many problems and disorders, which can cause weakness, pain, and loss of movement or paralysis. Some of these muscle Manuscript received December 12, 2012; accepted March 16, 2013. Date of publication March 29, 2013; date of current version May 29, 2013. The associate editor coordinating the review of this paper and approving it for publication was Prof. Roozbeh Jafari. The authors are with the AI-Rehab Research Group, Universiti Malaysia Perlis, Perlis 02600, Malaysia (e-mail: anamulislam.phd@gmail.com; kenneth@unimap.edu.my; badli@unimap.edu.my; ahamed1557@hotmail.com; asrafbabu@hotmail.com). Digital Object Identifier 10.1109/JSEN.2013.2255982 injuries may be very painful and of long duration. Therefore, a proper diagnosis is essential in order to identify and treat these muscle disorders. Consequently, many researchers have continued to explore suitable techniques, which include elec- tromyogram (EMG) [1], [2], myokinetic (MK) therapy [3], sonomyogram (SMG) [4], [5], tensiomyogram (TMG) [6], [7], and mechanomyogram (MMG) [8], [9]. Although the widely used EMG has attracted attention for some decades as a reliable supporting tool for the assess- ment of the skeletal muscles, MMG has been proposed as another tool to study muscle mechanical activity [10]. MMG (also termed as acousticmyography, soundmyography, phon- omyography, vibromyography, and acceleromyography; but in 1995 CIBA Foundation Symposium suggested a common term – MMG) records and quantifies the low-frequency lateral oscillations of active skeletal muscle fibers [11]–[13]. These oscillations which are generated by electrical stimulation rely on belly thickening [12], length of muscles [13], pressure waves [14], as well as muscle fiber recruitment [11], [12]. However, these oscillations reflect the mechanical counterpart of the motor unit (MU) electrical activity as measured by EMG [15], and these waves oscillate as discrete bursts rather than continuous tones [16]. The frequency content in the MMG signal is closely related to the resonant frequency of the muscle [17], which in turn is affected by muscle stiff- ness [13], [17]. During almost all voluntary muscle actions, MMG reflects the summation of the dimensional changes in the fibers of each nonlinearly recruited MU [18], [19]. Recently, Beck [20] abstracted that spike-triggered averaging techniques during voluntary contraction allows the activities from individual MUs to form MMG signal and the ampli- tude as well as frequency content of the signals produced by these MUs were influenced by the muscle morphology. However, Cescon et al. [21] found that MMG signals gener- ated by a single MU propagates in the transverse direction but not in the longitudinal direction at the location of the sensor with respect to muscle fibers. Furthermore, Archer and Sabra [22] claimed that MMG signals propagate in the lon- gitudinal direction if their frequency is greater than 20 Hz but this nature changes to transverse when the frequency is less than 20 Hz. Again, Archer et al. [23] found that MMG signals mainly propagate along the longitudinal direction of the muscle fibers at frequencies greater than 25 Hz. These works suggest and emphasize the need to further validate the characteristics of the different sources of MMG signals. 1530-437X/$31.00 © 2013 IEEE