Frequency Modulation based Resistive Sensing for Wearable Galvanic Skin Response

This paper presents a frequency modulation based readout circuit for the measurement of skin conductance or resistance. A charge pump based frequency-to-voltage converter circuit with adjustable sensitivity is used to convert the frequency shifts due to skin resistance changes into voltage variations. The readout circuit improves the measurement accuracy and artifact rejection in measurements of the galvanic skin response on the skin, and can be integrated with wearable physiological monitoring systems. The readout circuit is designed and fabricated using the UMC 0.18 μm CMOS technology. It occupies an area of 0.18 mm and consumes 11.7 mW.


INTRODUCTION
For monitoring health, activity, mobility, and bio-potential continuously, extensive research has been performed in the field of wearable biosensors and physiological monitoring systems [1,2,3,4,5].The major vital signs are electrocardiogram (ECG) for electrical activity of cardiac, heart sounds, electromyogram (EMG) for electrical activity of a muscle, electroencephalogram (EEG) for electrical activity of brain, electrical conductivity of skin, blood glucose, respiration rate, mobility for limb movement, and skin temperature [2,4].The sensors and low-power readout circuits with wireless are integrated to develop wearable physiological monitoring systems.
The galvanic skin response (GSR) is an indicator of sympathetic and parasympathetic nervous system activity that affect the skin conductivity due to the activity of sweat glands [4,5].Since the sweating is controlled by sympathetic and parasympathetic nervous system, skin conductance is a means for the measurement of sympathetic and emotional responses [6,7,8,9].The sympathetic response due to excitations increases sweat in sweat glands that results in parallel low-resistive pathways.As a result, the conductivity of the skin increases.The parasympathetic response due to relaxation diminishes the low-resistive pathways, and therefore, decreases the conductivity of the skin.The electrodes are positioned on the fingers and/or palm.
A readout circuit is needed to covert the skin conductance or resistance to measurable voltage levels.To measure the skin resistance, a constant current is applied across the electrodes, or a constant voltage is applied across a known resistor and the electrodes in series [1].Amplifier based readout circuit is also utilized to measure skin resistance.These circuits provide amplified voltage change proportional to the skin resistance change [5].Various noise sources such as flicker noise (1/f noise), thermal noise, amplifier noise, and substrate-noise coupling minimize the sensor readout range and resolution of these circuits [10].
In this paper, a frequency modulation based readout circuit is used for detecting galvanic skin resistance.A frequencyto-voltage converter (FVC) is used to convert the frequency shifts due to skin resistance into voltage change with adjustable sensitivity [11].The feedback control system minimizes the overall power consumption and presents a low power alternative for resistive readout circuits.This circuit can be integrated with wearable physiological monitoring systems.The readout circuit for galvanic skin resistance is presented in Section II.The experimental results are discussed in Section III.Section IV presents a summary of this paper.

INTERFACE CIRCUIT
The block diagram of the readout system for skin resistance is shown in Fig. 1.A differential crossed-coupled VCO is implemented to convert skin resistance changes (∆Rsen) into frequency variation (∆Fvco).A sine-to-square (STS) converter circuit converts ∆Fvco from the VCO into time-  period variation (∆Tvco).A frequency-to-voltage converter (FVC) circuit is designed to convert ∆Tvco as the voltage changes (∆V ) [11].
The schematic of a differential crossed coupled VCO employing an inversion-mode MOS (I-MOS) varactor is illustrated in Fig. 2(a).The I-MOS varactor, C0 [12] and inductor, L0 act as the LC-tank of the VCO (Fig. 2(b)).The C0 is dependent on the source-drain voltage, V ctrl .The effective capacitance is modeled by C0(V ctrl ) = Cmin + CV (V ctrl ) for V ctrl > 0 condition [12].Here, Cmin is the minimum fixed capacitance of the varactor and CV is the additive capacitance component that depends on V ctrl .The skin resistance (RSEN ) is in series with a reference resistor (RREF ) to produce a varying control voltage (V ctrl ) with respect to RSEN .The varying voltage, V ctrl , resulting from skin resistance is translated to frequency variation using the VCO.The VCO oscillation frequency can be approximated for this configuration as Fvco = 1/2π √ L0 (Cmin + CV (V ctrl )).The time period of the oscillator for V ctrl can be expressed as The FVC block comprising of the logic controller block and a charge pump (CP) circuit, is illustrated in Fig. 3(a).An external control voltage (VBD) and a feedback voltage (VF V C ) from the CP circuit provide the control to translate the The FVC output voltage for the varactor capacitance and time-delay feedback can be approximated as where Qc = PcT0/2Cmin, Pc = I ch /C1 − 1/R2C2, and TID is the fixed delay from inverter.Here, KF D and KBD are the slopes of RC delay circuits.From (2), the slope of the VF V C for the variation of ∆CV (V ctrl ) is Qc/ (1 + PcKF D ).Since, Pc and Qc are proportional to I ch , the slope of VF V C is controlled by I ch .The VCO and FVC circuits are fabricated in the UMC 0.18 µm CMOS process and occupy an active area of 0.18 mm 2 .

EXPERIMENTAL RESULTS
The performance of the FVC circuit is verified by applying a set of controlling parameters (I ch and VBD) for the variation of off-chip (SMD) resistors.Then, the readout circuit with FVC is evaluated for the measurement of GSR under the influence of controlling parameters.
The experimental measurements of the FVC output voltage for the variation of V ctrl is shown in Fig. 4. The VCO circuit (L0=15 nH) modulates from 203.8 to 195.5 MHz for the variation of V ctrl from 1.0 to 1.45 V.The V ctrl that changes according to the SMD resistors, is converted to frequency variations with a sensitivity of 18.44 MHz/V.The FVC translates the frequency variations to voltage variations with a sensitivity of 693.1 mV/V for I ch of 24.53 µA.
If I ch is increased to 41.7 µA, the sensitivity increases to 730 mV/V, whereas the sensitivity decreases to 545.1 mV/V for a decreased I ch of 6.76 µA.The VBD is fixed at 0.4 V.As a result, the sensitivity of the output voltage is also controlled by the I ch .
The GSR is measured by using two dry Ag-AgCl electrodes placed at the thenar and hypothenar positions of the palm, as shown in Fig. 5.The area of the each electrode is 1 cm 2 .The range of galvanic skin resistance is 0 to 100 KΩ.The readout circuit is connected to measure skin resistance, as The time-domain plot of the resistance (RSEN ) from GSR response, input control voltage (V ctrl ) from resistor divider circuit, and FVC output voltage (VF V C ) for muscle and painful stimulus are shown in Fig. 6.The acquired signals are sampled at 1000 samples/s and digitized at 12-bit resolution.The muscle contractions are performed at 6th and 39th minute of the experiment.The muscle relaxations are performed after 1 min of muscle contractions.The painful stimulus of pinprick is performed at 16th minute of the experiment.The muscle contraction and painful stimulus increases sweat secretion that in turn decreases the resistance and increases the input control voltage.The reverse events decreases the input control voltage.The input control voltage from GSR response produces noises due to the artifacts caused by motion, skin-electrode contact, and sweating.The FVC output voltage, which provides lower noises and higher temporal correlation, improves the detection of high-amplitude events compared to the simple resistor divider circuit.It also provides higher sensitivity to smaller  amplitude events that were ignored in standard GSR recording.

CONCLUSIONS
A low noise, low power, and high resolution galvanic skin resistance readout circuit is presented.The circuit is designed using the UMC 0.18 µm technology.The changes of resistance are translated into VCO oscillation frequencies and pulse-width of control signals, which are detected using the FVC circuit as voltage outputs.The circuit has been evaluated for the events of muscle and painful stimulus.This circuit provides improved measurement accuracy and artifact rejection in measurements, higher sensitivity, and lower noise level compared to the resistor divider circuit.

Figure 1 :
Figure 1: Block diagram of the proposed resistive sensor interface system.

Figure 2 :
Figure 2: (a) Schematic of crossed coupled differential VCO circuit with a configurable LC-tank circuit.(b) The schematic of the LC-tank circuit for the resistive sensor system where sensor (RSEN ) is in series with a reference resistor (RREF ) to produce a varying control voltage (V ctrl ) with respect to RSEN .

Figure 3 :
Figure 3: (a) Block diagram of the FVC circuit with feedback charge pump.(b) Schematic of logic controller circuit.(c) Schematic of the charge pump circuit.

Figure 4 :
Figure 4: Experimental measurement of FVC output voltage and VCO oscillation frequency for the variation of varactor control voltage.

Figure 5 :
Figure 5: Experimental setup for the GSR measurement.

Figure 6 :
Figure 6: Time-domain plot for muscle and painful stimulus of (a) the skin resistance, (b) input control voltage from resistor divider circuit, and (c) FVC output voltage.

Figure 7 :
Figure 7: Measured spectral density of the FVC output voltage and the input control voltage.

Figure 7
Figure 7 depicts the measured spectral density of the FVC output voltage (VF V C ) and the input control voltage (V ctrl ) of the readout circuit.The signals were recorded for each 40 minutes.The 1/f corner frequency at the output of the circuit is 100 Hz , which is lower (around 500 Hz) than the input of the circuit.The measured thermal noise floor is 47.84 µV/ √ (Hz) at the input of the readout circuit.A decrease in the thermal noise floor of 8.14 µV/ √ (Hz) is observed at the output of the readout circuit.Table 1 show a comparison for the GSR measurement techniques with previous techniques in the literature.

Table 1 : Comparison of the interface circuits for GSR measurements.
Table 1 show a comparison for the GSR measurement techniques with previous techniques in the literature.