Fig 13 - uploaded by Venkata Rajesh Pamula
Content may be subject to copyright.
Source publication
A compressive sampling (CS) photoplethysmographic (PPG) readout with embedded feature extraction to estimate heart rate (HR) directly from compressively sampled data is presented. It integrates a low-power analog front end together with a digital back end to perform feature extraction to estimate the average HR over a 4 s interval directly from com...
Context in source publication
Context 1
... val is calculated by accumulating the samples as they arrive at the input of the DBE and dividing by 4 × f s,CS , where f s,CS is the average sampling frequency given by (4). The division process is performed through nonrestoring divide algorithm. The mean subtracted samples are then fed into an eight-way multiply-accumulate (MAC) unit, shown in Fig. 13, which per- forms the acceleration of the matrix multiplication operation in (12). Of eight MAC units, four are assigned to accelerate the multiplication with cosine coefficients, while the rest acceler- ate the sine coefficient multiplications, thereby requiring 8192 clock cycles for the FEU to compute the LSP coefficients of the 4 s ...
Citations
... The LED driver power reduction of our sensor is 85.91%, surpassing the 74.72%-84.76% reported in [11] and [10], respectively. While [29] attains the highest LED driver power reduction of 96.41%, it does so at the cost of an HR MAE as high as 10 bpm. The AFE power reduction of our sensor is 89.31%, surpassing the 73.33%-86.97% in [10] and [11], respectively. ...
... In contrast, static W -based PPG sensors, as in [10], uses a predefined static W for all cardiac cycles, and [11] uses a W extracted from a few cardiac cycles at the beginning of the measurement and uses it for all the next cardiac cycles, increases the W by a fixed value if a cardiac peak is missed, and then uses this new static W for the next cycles again, limiting their performance to fixed measurement conditions without MA. Our sensor also achieves an FoM LEDDriver of 171.82 (%/bpm), significantly higher than the 74.72, 16.95, and 9.641 (%/bpm) in [10], [11], and [29], respectively. The FoM AFE of our sensor is 178.62 (%/bpm), surpassing 73.33 and 17.39 (%/bpm) in [10] and [11], respectively. ...
Photoplethysmography (PPG) sensors provide precise vital sign measurements. However, conventional PPG signal detection requires continuous light-emitting diodes, resulting in significant power consumption. To cancel DC components, ambient light, and motion artifacts, PPG sensors incorporate significant signal processing circuitry. This circuitry is critical to cancel interference, increase dynamic range, and improve signal-to-noise ratio, which further increases power consumption. This paper presents a novel PPG sensor that adaptively predicts and samples only the peaks and valleys (PAVs), i.e., the maximum and minimum amplitudes, of the PPG signal. Using the predicted PAVs, the sensor extracts vital signs, including heart rate (HR), HR variability (HRV), oxygen saturation (SpO
<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub>
), systolic, diastolic, and cardiac cycles, while reducing power consumption by over 87%. Our sampling scheme shows a low mean absolute error (MAE) of 5.76 beats per minute (bpm) when tested on the PPG-DaLiA dataset with data from 15 subjects. Moreover, in a prototype test with five subjects, our sensor was used on various body parts, including the finger, wrist, arm, chest, neck, earlobe, and forehead, during activities such as walking, typing, and sitting. Achieved MAEs are <0.5 bpm for HR, <1% for SpO
<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub>
, and <3 ms for HRV, systolic, diastolic, and cardiac cycles, across all subjects.
... Based on a comprehensive noise analysis of TIA topologies in [34], RC-TIAs often need ∼8 times less LED power than C-TIAs to achieve the same SNR at the same bandwidth (BW). By having both resistors and capacitors, RC-TIAs also reduce the effect of the PD capacitance, C PD , on the noise transfer function, which is approximately equal to (1 + C PD /C F ), where C F is the feedback capacitor of the RC-TIA [35]. ...
... The flicker noise associated with the ambient light cancellation IDACs is also removed using correlated double sampling, taking an ambient sample when the LEDs are OFF and subtracting it from a PPG sample when they are ON [53] at the cost of more power [10]. This work achieves an input-referred noise of 35 In [36], a dual-wavelength RC-TIA-based PPG sensor is designed with a total power of 336.6 µW at a duty cycle of 0.7%, as shown in Fig. 8. To reduce power consumption, this work uses a modulation, filtering, and decimation process to remove time-varying interferers, achieving an ambient light attenuation of 87 dB without sacrificing much power. ...
... C. Compressive Sampling 1) Resistive and Capacitive TIAs: In [35], a compressively sampled RC-TIA-based PPG sensor is designed to predict HR in real time. In compressive sampling, the measured data are traditionally sent to a base station to reconstruct the compressed signal [70] to save energy at the cost of real-time monitoring. ...
PPG sensors are used to accurately, instantaneously, and non-invasively measure vital signs to provide a real-time indication of overall physical health and long-term well-being. Achieving long-term continuous monitoring is an important requirement to increase user safety and diagnostic accuracy. PPG sensors need a light-emitting diode (LED) with sufficient output power to detect the PPG signal, which consumes tens of mW. On the other hand, low AC/DC ratios of <0.1-4%, ambient light, motion artifacts, and semiconductor noise greatly affect the signal-to-noise ratio (SNR), dynamic range (DR), and signal quality. Specialized circuit blocks are needed to cancel these interferences, further increasing power consumption. Several ultra-low-power designs, circuit techniques, and sampling schemes have been proposed in the literature to extend PPG sensors’ lifetime. This paper reviews, analyzes, and critiques these solutions to provide designers with comprehensive design considerations for achieving ultra-low power consumption while achieving the required SNR and DR in a PPG sensor design.
... However, since there is no existing relationship between p and v, the backpropagation step for v → p must be newly designed. Pamula et al. [50] proved that the rPPG signal can approximately satisfy the characteristics of a sinusoidal signal. Therefore, we assume that x s satisfies the sine curve. ...
Noncontact heart rate (HR) measurement is a very important trend in clinical medicine. Recently, a variety of deep networks have been applied to estimate HRs from facial videos. However, due to limited data resources and poor parameter optimization, few existing models have achieved incredible performance in complicated scenarios, such as those with illumination changes, different skin tones, and facial motion. To address these challenges, this paper proposes a novel multi-stage deep network (MSDN) that can decentralize the learnable parameters into different stages to reduce the difficulty of learning through multiple training steps. Specifically, the proposed network consists of three stages in an end-to-end way. In the first stage, an HR-aware feature extractor utilizes the next convolutional neural network (ConvNeXt) embedded with a newly designed bandpass filter as its backbone to extract spatial-temporal features for determining heart rate changes. Moreover, pseudolabels are generated to make the features compatible with illumination, motion, and color variance. In the second stage, various modules, including singular value decomposition (SVD) pooling and enhanced difference convolution modules, are then designed and combined with a Transformer encoder to construct a feature-compressed remote photoplethysmography (rPPG) generator. In the third stage, an HR estimator with an interbeat interval analyzer and a 1D filter is newly designed for HR estimation. Extensive experiments are performed on three publicly available databases (i.e., VIPL-HR, COHFACE, and PURE), and the results demonstrate the effectiveness of the proposed method through ablation studies and comparison experiments with state-of-the-art (SOTA) methods.
... The technique significantly reduced the readout power, but the LEDs still dominated the sensor power consumption. Compressive sampling was first introduced in [19] to exploit the sparse nature of the PPG signal and thus save power by reducing the number of sample points. Despite excellent results for HR estimation from the compressively sampled data, SpO 2 measurements required full reconstruction of the PPG waveform out of randomly selected samples, demanding up to 10 mW of processing power [19]. ...
... Compressive sampling was first introduced in [19] to exploit the sparse nature of the PPG signal and thus save power by reducing the number of sample points. Despite excellent results for HR estimation from the compressively sampled data, SpO 2 measurements required full reconstruction of the PPG waveform out of randomly selected samples, demanding up to 10 mW of processing power [19]. [20] presented the heart-beat-locked-loop technique to make the sensor lock to the PPG signal period and selectively sample the PPG peaks to report HR data. ...
Photoplethysmography (PPG) is an attractive method to acquire vital signs such as heart rate and blood oxygenation and is frequently used in clinical and at-home settings. Continuous operation of health monitoring devices demands a low power sensor that does not restrict the device battery life. Silicon photodiodes (PD) and LEDs are commonly used as the interface devices in PPG sensors; however, using of flexible organic devices can enhance the sensor conformality and reduce the cost of fabrication. In most PPG sensors, most of system power consumption is concentrated in powering LEDs, traditionally consuming mWs. Using organic devices further increases this power demand since these devices exhibit larger parasitic capacitances and typically need higher drive voltages. This work presents a sensor IC for continuous SpO$_2$ and HR monitoring that features an on-chip reconstruction-free sparse sampling algorithm to reduce the overall system power consumption by $\sim$70\% while maintaining the accuracy of the output information. The designed frontend is compatible with a wide range of devices from silicon PDs to organic PDs with parasitic capacitances up to 10 nF. Implemented in a 40 nm HV CMOS process, the chip occupies 2.43 mm$^2$ and consumes 49.7 $\mu$W and 15.2 $\mu$W of power in continuous and sparse sampling modes respectively. The performance of the sensor IC has been verified \textit{in vivo} with both types of devices and the results are compared against a clinical grade reference. Less than 1 bpm and 1\% mean absolute errors were achieved in both continuous and sparse modes of operation.
... Then, the static component can be compensated by the analogous DAC current adjusted by the MCU. Hence, the remaining AC signal can be amplified and the overall DR of the readout channel can be improved [19]. However, because of the additional DAC in parallel with PD, the input-referred low frequency noise will be increased in such a way [20]. ...
... So that low frequency noise in the different PPG signal data is effectively attenuated. The systematic CDS [17][18][19][20] occurs at the MCU to obtain the low noise PPG signal data. The CDS operation is a technique that adopts two timing phases to sample signals and fixed pattern noise respectively, and one phase to sample the signals with noise and another to sample the fixed pattern noise. ...
... In recent years, significant efforts have focused on developing integrated recording systems that feature low power consumption, wide DR, and high resolution for powerefficient, high-fidelity recording of the small-amplitude, ac, and PPG signals in the presence of ambient interference light-induced components [17][18][19][20][21][22][23][24]. The PD in the RX not only senses the LED modulated PPG signal, but also senses a significant amount of ambient interference current signals . ...
This paper presents a low-noise analog front end (AFE) with interstage systematically ambient interference cancellation for a pulse oximeter, which is suitable for clinical oxygen saturation (SPO2) detection with a low perfusion index. The fully differential implementation is adopted to have a better rejection performance of common mode interference and noise. Firstly, the proposed interstage systematically ambient interference cancellation is placed in the two gain stages to decrease low frequency noise in the bandwidth of interest (0.1–20 Hz), so that the larger signal-to-noise ratio (SNR) can be achieved to increase the detection accuracy of this system. Secondly, due to the additional gain stage compared with traditional implementation, the current-reuse OTA is employed to have better noise and power efficiency. Thirdly, the auto zero technique is utilized in the LED Driver to decrease the offset voltage and acquire a larger dynamic range (DR) in the low frequency bandwidth of interest. This PPG AFE chip is designed and fabricated in a 180 nm standard CMOS process. The receiver (RX) of this AFE consumes 220 μW from a 1.8 V supply, and the power consumption of the transmitter (TX) is 60 μW from a 3 V supply. The measurement results show that the input-referred noise current of 2.3 pA/sqrt(Hz) is achieved in RX and 110 dB peak DR is obtained in TX.
... Initial work indicates that PPG signal quality is highest during sleep when movement and ambient light are minimal, supporting the approach of only acquiring PPG signals during periods of low activity [29]. Other approaches to reduce power consumption include: 1) delaying the next measurement after one of low quality [86]; 2) using compressive sampling to reduce the sampling frequency while still being able to accurately obtain information from PPG pulse waves [270], [271]; and 3) using windowed sampling to only sample portions of interest of the PPG pulse wave (such as systolic peaks) [272]. ...
Smart wearables provide an opportunity to monitor health in daily life and are emerging as potential tools for detecting cardiovascular disease (CVD). Wearables such as fitness bands and smartwatches routinely monitor the photoplethysmogram signal, an optical measure of the arterial pulse wave that is strongly influenced by the heart and blood vessels. In this survey, we summarize the fundamentals of wearable photoplethysmography and its analysis, identify its potential clinical applications, and outline pressing directions for future research in order to realize its full potential for tackling CVD.
... One of the approach to reduce transmission or/and storage energy consumption is the data compression by using a lightweight compression method [16]- [18]. In the past studies, compressed sampling technique was explored for reducing energy consumption at the sensing node [9], [27]- [29] that reconstruct the original signal from a few measurements which is lesser than number of samples obtained with Nyquist-sampling rate. Another approach for avoiding transmission of vast data is enabling automatic on-device parameter extraction or edge-device feature extraction by exploring lightweight and fast signal processing techniques [10], [11]. ...
... In the past studies, baseline wander is characterized as a single or mixture of low-frequency components having frequencies below 1 Hz with time varying amplitudes. The baseline artifact corrupted or respiratory-induced PPG signal can be modelled as (27) and can be rewritten as the baseline wander artifacts due to the movements of sensors or/and subjects. Based on the hypothesis that baseline wander does not carry clinical information related to the most of PPG features and pulse parameters, this study mainly focuses on the quantitative measure of pulsatile waveform morphological feature distortion due to the lossy processing techniques. ...
... In the literature, compressive sampling (CS) based data reduction methods were reported in the analog-domain [9], [27]- [29]. The digital CS method was explored due to its simplicity in real-time implementation on the resource-constrained device [88], [89] which process discrete signal using the CS theory for data reduction. ...
Real-time photoplethysmogram (PPG) denoising and data compression has become most essential requirements for accurately measuring vital parameters and efficient data transmission but that may introduce different kinds of waveform distortions due to the lossy processing techniques. Subjective quality assessment tests are the most reliable way to assess the quality, but they are time expensive and also cannot be incorporated with quality-driven compression mechanism. Thus, finding a best objective distortion measure is highly demanded for automatically evaluating quality of reconstructed PPG signal that must be subjectively meaningful and simple. In this paper, we present four types of objective distortion measures and evaluate their performance in terms of quality prediction accuracy, Pearson correlation coefficient and computational time. The performance evaluation is performed on different kinds of PPG waveform distortions introduced by the predictive coding, compressed sampling, discrete cosine transform and discrete wavelet transform. On the normal and abnormal PPG signals taken from five standard databases, evaluation results showed that different subjective quality evaluation groups (5-point, 3-point and 2-point rating scale) had different best objective distortion measures in terms of prediction accuracy and Pearson correlation coefficient. Moreover, selection of a best objective distortion measure depends upon type of PPG features that need to be preserved in the reconstructed signal.
... It describes the integration of an application-specific integrated circuit (ASIC) custom-designed for low-power PPG recording. The ASIC was designed as a part of a collaboration between our research group, iBionicS at NC State University, Raleigh, NC, USA, and imec at Leuven, Belgium to implement a low-power signal acquisition technique called compressed sensing (CS) [92]. This work assembles this CS PPG ASIC into a wearable wristband form factor and performs preliminary validation experiments to compare with commercial devices [93]. ...
... The joint team designed and fabricated an ASIC in 0.18 µm CMOS process that consumes 172 µW power to extract heart rate from the randomly sampled PPG signal [92]. This ASIC is capable of extracting the HR data from the compressed PPG signal with an effective sampling rate of only 4 Hz. ...
... Hence, the second stage in the readout circuit is a switched integrator (SI) which also adds extra gain to the signal path tunable by programming the feedback capacitor (C int ) [92]. ...
The relentless pursuit of financial efficiency has encouraged the development of intensive animal management systems, where the care of the animal is sometimes compromised. As the physical or emotional stress on the animals summons the conscience of the consumers, the public's interest in animal welfare is continuing to rise. While several qualitative and quantitative measures are used to assess the long-term welfare of an animal, the physiological and behavioral states of the animals are the only quantifiable measures of the short-term responses of animal welfare. Moreover, studying the vital signs [e.g., heart rate (HR), breathing rate (BR), blood pressure (BP), core body temperature, etc.] and behavioral traits of freely moving animals can provide significant insights to veterinarians, animal researchers, and biomedical engineers. Monitoring of animals is also necessary for the pharmaceutical industries, where the safety and efficacy of human drugs are tested on animal models. Wireless sensor systems attached to individual animals can provide specific physio-behavioral information about each animal continuously. However, an externally attached device on a freely moving animal would have unfavorable impacts on its natural behavior and comfort. Moreover, the recordings from a wearable sensor would suffer from the obstruction created by the layer of skin and fur. An implantable system, on the other hand, can avoid the difficulties related to the attachment of sensors to the animal and can be minimally obtrusive, depending on the size of the implant. In this research, a subcutaneously injectable implant equipped with several sensing capabilities is developed using commercial-off-the-shelf components.
First, the transparently encapsulated implant includes a biophotonic front-end circuit that can acquire photoplethysmography (PPG) signals. The designed system successfully recorded PPG signals using light sources of different wavelengths from rats and chickens during \textit{in vivo} experiments. As PPG systems are highly power-consuming, a low-power custom-integrated PPG front-end circuit has been validated by developing a wearable wristband for humans that has the potential to reduce the implant’s battery usage in the future. Second, the developed system is capable of biopotential (electrocardiography or ECG) and bioimpedance (BIOZ) measurements that can provide deeper insight into the cardiovascular system. Despite the difficulties of interfacing conductive electrodes in implants, two techniques for manufacturing electrode surfaces on the implant are proposed, and the accuracy of the system is validated with a commercial ECG amplifier during the in-vivo experiments. The combination of this biophotonic and bioelectric sensing would enable the estimation of HR, BR, oxygen saturation in the blood (pulse oximetry), pulse transit time (PTT) which is correlated with BP, tissue hydration level, etc. Third, a temperature sensor has been added to read the core body temperature, which has been validated using an in-vitro setup. Lastly, an inertial measurement unit (IMU) that integrates an accelerometer and a magnetometer are included in the system. Accelerometry can track various micro and macro activities by classifying the tri-axial data, whereas magnetometry can register an animal's physical orientation.
All these sensor electronics, along with a wireless microcontroller and a pin-type battery, are coated with biocompatible materials and packaged into a capsule-shaped cylinder with a diameter of 4 mm. This miniaturized implant fits into a commercially available injector (similar to the ones used for RFID tags) and allows for an easier injection method avoiding any surgical procedure on the animal. The contribution of this research includes the design and development of the implantable system, optimization of the hardware and software to reduce the power consumption, packaging innovations to accommodate electrical interfaces within the injectable form factor, and the in-vivo animal experiments for the validation of individual sensors.
... First of all, the raw PPG signals are bandpass-filtered from 0.5 Hz to 5 Hz before any further processing to exclude frequency components that are not physically possible correlated with the PPG data [37][38][39][40][41]. For this step, we used minimum-order filters with a stopband attenuation of 60 dB and compensation for the delay introduced by the filter. ...
The photoplethysmographic (PPG) signal is an unobtrusive blood pulsewave measure that has recently gained popularity in the context of the Internet of Things. Even though it is commonly used for heart rate detection, it has been lately employed on multimodal health and wellness monitoring applications. Unfortunately, this signal is prone to motion artifacts, making it almost useless in all situations where a person is not entirely at rest. To overcome this issue, we propose SPARE, a spectral peak recovery algorithm for PPG signals pulsewave reconstruction. Our solution exploits the local semiperiodicity of the pulsewave signal, together with the information about the cardiac rhythm provided by an available simultaneous ECG, to reconstruct its full waveform, even when affected by strong artifacts. The developed algorithm builds on state-of-the-art signal decomposition methods, and integrates novel techniques for signal reconstruction. Experimental results are reported both in the case of PPG signals acquired during physical activity and at rest, but corrupted in a systematic way by synthetic noise. The full PPG waveform reconstruction enables the identification of several health-related features from the signal, showing an improvement of up to 65% in the detection of different biomarkers from PPG signals affected by noise.
... The complexity of the next stage of the pre-processing circuit, i.e., signal acquisition incorporating interface, signal-conditioning, amplification, filtering, and analog to digital (A/D) conversion determines its power consumption, and application-specific integrated circuits (ASICs) can be optimized to consume as low as tens to hundreds of µW of power. For example, an ASIC integrating low-power analog front end and a digital back end for PPG monitoring was shown to consume~172 µW of energy [34]. As compared to the sensing and pre-processing circuits, the data processing (microcontroller units (MCUs)), storage (memory), and transmission (Wi-Fi, Bluetooth) block of the system architecture is the most power-hungry element and provides the largest obstacle towards self-powered, end-to-end solutions. ...
The use of rapid point-of-care (PoC) diagnostics in conjunction with physiological signal monitoring has seen tremendous progress in their availability and uptake, particularly in low- and middle-income countries (LMICs). However, to truly overcome infrastructural and resource constraints, there is an urgent need for self-powered devices which can enable on-demand and/or continuous monitoring of patients. The past decade has seen the rapid rise of triboelectric nanogenerators (TENGs) as the choice for high-efficiency energy harvesting for developing self-powered systems as well as for use as sensors. This review provides an overview of the current state of the art of such wearable sensors and end-to-end solutions for physiological and biomarker monitoring. We further discuss the current constraints and bottlenecks of these devices and systems and provide an outlook on the development of TENG-enabled PoC/monitoring devices that could eventually meet criteria formulated specifically for use in LMICs.
