Compression of PPG Signal Through Joint Technique of Auto-Encoder and Feature Selection

Compression of PPG Signal Through Joint Technique of Auto-Encoder and Feature Selection

Sunil Kumar K. N., Shiva Shankar, Keshavamurthy
DOI: 10.4018/IJHISI.20211001.oa23
Article PDF Download
Open access articles are freely available for download

Abstract

PPG signal utilize the light-based method to sense the blood-flow-rate as controlled by the actions of heart’s pumping. It is extensively utilized in the healthcare with application ranging from the pulse oximetry in the serious care units to the heart rate (HR) measurement in the wearable devices. This paper introduces the algorithm known as PPGC-AE-FS (PPG-Signal Compression using Auto-Encoder and Feature Selection) that is the combined generative method, which incorporates FS and AE together. At the end, our introduced algorithm can differentiate the task as relevant units through not relevant task to get very effective feature for the classification task. Our method not only accomplishes the FS on the learned level of higher feature, but also endows the AE to construct the discriminative units. Our experimental outcomes on many benchmarks that demonstrate our model is much better than existing methods.
Article Preview
Top

Introduction

The monitoring of BP (Blood Pressure) that reflects the stress, arteriosclerosis and so on, which is very effective for the management of health. It is very essential to check the variation of BP that changes according to state whether it is active or in the rest and with the peripheral environment. The traditional measurement of BP must fasten to the upper arm of persons. Then, the person can be applied the burden from fastening the cuff and that cannot measure the BP. The less measurement of BP technique is needed for daylong the BP monitoring. The signal of pulse-wave has very strong correlation with the cardiovascular events (Blacher et al., 1999) that has been utilized to the estimation of cuff-less BP. The signal of pulse wave can be easily measured and noninvasively by the help of PPG (Photoplethysmography) sensor. The PPG sensors utilize the light-based method to sense the bold-flow-rate as controlled by the actions of heart’s pumping. It is extensively utilized in the healthcare with application ranging from the pulse oximetry in the serious care units to the heart rate (HR) measurement in the wearable devices. It is very popular in the global healthcare beyond the measurement of HR because of providing the patients information (Tison et al., 2018). For instance, the author (Lin et al., 2017), introduced the method of BP estimation based PTT (Pulse Transmit Time) that is the time variance among PPG and ECG (electrocardiogram). Equally, estimated the continuous BP via PPG and PTT based DL (Deep learning). Anyways, these techniques are impractical due to utilizing more sensors as PPG and ECG.

Although, many works indicate the advantageous prospect for improvement of PPG signal based technology, PGG is challenged via its vulnerability to different types of source noise. Whereas, PPG needs roust pre-processing phase to eliminate the stationary noise like power line-interference and non-stationary, which is caused by the movement like voluntary and breathing motion, and resulting variation in contact among device and skin. As underlying the essential structure of skin, that defines amplitudes of measured skin, the PPG sensors can be preceded by sudden movement and the automatic gain controller or transformations in the contact that can cause the temporary signal variations to flat lining and saturations. As more practical and wearable technique of the BP estimation, some of the techniques utilizing the PPG sensor that has been introduced. In Nishio et al., (2016), introduced the estimation of BP via PPG signal utilizing multi-task of the Gaussian processes. In Kishimoto et al., (2015), the author have estimated the BP by removing the elapsed time and wave-height from developing the point of pulse-wave to the features point in APG (accelerated plethysmogram) and PPG and utilized them as the feature of ML technique. Furthermore, in Kishimoto et al., (2015), utilized the non-linear regression technique to evaluate the BP. Anyways, the BP is not estimated precisely and put into the practical utilization in any case.

The acquisition of PPG units generates about 415KB data in one-min at rate of 500HZ sampling with the resolution of 16-bits (Karlen et al., 2013) that is very larger. Moreover, the bandwidth of PPG signal can be assumed at the range of 0.05-15 Hz (PhysioNet, n.d.). Henceforth, the signal of PPG signal at 500Hz, which is redundant. Therefore, the efficient compression algorithm of PGG is needed to complete with higher developing demand of storage rate. However, some researchers have been compress the digitalized PPG signal in recent times as well as their performance cannot satisfactory. In Gupta, (2015), introduced the compression algorithm of lossless PPG based HF (Huffman coding) and second difference but their performance is very poor to be utilized in the real-time application. So, the method of delta modulation is utilized in Chong, (2013) to compress the signal of PPG. This compression method could only be the beneficial, is signal can be samples at higher rate. The signal of PPG can be recorded at the sample rate of 1-KHz that is very redundant.

Complete Article List

Search this Journal:
Reset
Volume 19: 1 Issue (2024)
Volume 18: 1 Issue (2023)
Volume 17: 2 Issues (2022)
Volume 16: 4 Issues (2021)
Volume 15: 4 Issues (2020)
Volume 14: 4 Issues (2019)
Volume 13: 4 Issues (2018)
Volume 12: 4 Issues (2017)
Volume 11: 4 Issues (2016)
Volume 10: 4 Issues (2015)
Volume 9: 4 Issues (2014)
Volume 8: 4 Issues (2013)
Volume 7: 4 Issues (2012)
Volume 6: 4 Issues (2011)
Volume 5: 4 Issues (2010)
Volume 4: 4 Issues (2009)
Volume 3: 4 Issues (2008)
Volume 2: 4 Issues (2007)
Volume 1: 4 Issues (2006)
View Complete Journal Contents Listing