Digitally Controlled Feedback for DC Offset Cancellation in a Wearable Multichannel EMG Platform M. Tomasini, S. Benatti, F. Casamassima, B. Milosevic, S. Fateh, E. Farella and L. Benini Abstract— Wearable systems capable to capture vital signs allow the development of advanced medical applications. One notable example is the use of surface electromyography (EMG) to gather muscle activation potentials, in principle an easy input for prosthesis control. However, the acquisition of such signals is affected by high variability and ground loop problems. Moreover, the input impedance influenced in time by motion and perspiration determines an offset, which can be orders of magnitude higher than the signal of interest. We propose a wearable device equipped with a digitally controlled Analog Front End (AFE) for biopotentials acquisition with zero-offset. The proposed AFE solution has an internal Digital to Analog Converter (DAC) used to adjust independently the reference of each channel removing any DC offset. The analog integrated circuit is coupled with a microcontroller, which periodically estimates the offset and implements a closed loop feedback on the analog part. The proposed approach was tested on EMG signals acquired from 4 subjects while performing different activities and shows that the system correctly acquires signals with no DC offset. I. INTRODUCTION Recent technological advances and the application push from the consumer market have created a growing demand for wearable, low-cost and low-power devices with advanced bio-sensing capabilities. The processing of body-related po- tentials is becoming one of the main challenges in this field, leading towards efficient and unobtrusive applications. For this reason, the research on Analog-Front-Ends (AFEs) and on advanced energy efficient computing platforms is evolving both at silicon and at system level. One of the most interesting biopotentials is the surface EMG signal that allows to detect the muscular activity of the human body [1]. EMG is widely adopted in human machine interaction applications, such as prosthesis control [2], even if it is affected by high variability and ground loop problems. For this reasons, in prosthetics and medical applications, where a high-quality signal is required, expansive analog sensors are adopted. Ottobock sensors [3] represent the commercial solution for EMG acquisition for high-end pros- thetics, both in research and industrial applications. These sensors perform a full-analog signal conditioning based on a bandpass discrete filter, an instrumentation amplifier (IA) with a high gain stage and an offset cancellation feedback M.Tomasini, S.Benatti, F.Casamassima and L.Benini are with DEI, Unversity of Bologna, Bologna, Italy e-mail: {simone.benatti;filippo.casamassima;luca.benini}@unibo.it B.Milosevic and E.Farella are with E3DA, FBK, Trento, Italy. e-mail: {milosevic;efarella}@fbk.eu S.Fateh and L.Benini are with the IIS, ETHZ, Zurich, Switzerland. e-mail: {fateh;lbenini}@iis.ee.ethz.ch Manuscript received March 31, 2015; revised ... circuit that requires the use of a dedicated metal plate as the reference electrode for each sensor. An alternative approach, mostly used at research level, is based on passive low-cost electrodes, which requires to place a common reference electrode on the user’s body in a neutral position (e.g. on the elbow when collecting forearm signals). In this case, there is only one common reference, which is not adequate for all the channels. Each channel is therefore affected by an offset dependent on the input impedance and it is influenced by its position and the physical connection between the electrode and the skin. The input impedance changes every time we place the electrodes on the skin and it changes during an acquisition due to skin condition or electrode movement. With the common reference approach, a digital signal processing stage is required to obtain a well conditioned, zero mean differential signal, suitable for further analysis. Nevertheless, a correct offset removal is not possible if the signal is unbalanced and near the saturation voltage of the input amplifier. In the case of saturation, a digital offset removal on the acquired signal via mean subtraction is not possible without a loss in the signal content. The approach based on dedicated ASICs is becoming more promising for wearable biosignal acquisition platforms, prin- cipally for the reduction of costs and external components. The Texas Instrumets ADS1299 and the Analog Devices AD7194 are commercially available solutions, proposing AFEs with high CMRR, 24-bit resolution and a digital back- end suitable for integration with low power microcontrollers (MCUs). Nevertheless, the design of these components lacks flexibility because they are oriented to simplify the design of ECG systems with dedicated internal registers and circuits. The literature of research on ASIC design is rich of inspiring solutions for EMG and other biopotential signal acquisitions [4], [5], [6]. These works are focused on perfor- mance optimization in terms of CMRR and ADC resolution or bandwidth, but they do not consider the DC offset removal necessary for the signal processing in prosthetics. In fact, the offset due to input impedance and motion arti- facts can be orders of magnitude higher than the biomedical signal being recorded. This tends to saturate a high gain IA causing losses of information content of the EMG signal. To overcome these problems, in [7] a pair of servo loops are used, combining a 16-element DAC and a fine analog feedback to remove DC offset up to ±45mV. An alternative approach is AC coupling performed at the input of the IA [8]. In this work, we propose a digitally controlled AFE solution for EMG signal acquisition with zero-offset. The