22.4 A 172µW compressive sampling photoplethysmographic readout with embedded direct heart-rate and variability extraction from compressively sampled data

• Published in: 2016 IEEE International Solid-State Circuits Conference (ISSCC) pp. 386 - 387
• Main Authors: Rajesh, Pamula Venkata, Valero-Sarmiento, Jose Manuel, Long Yan, Bozkurt, Alper, Van Hoof, Chris, Van Helleputte, Nick, Yazicioglu, Refet Firat, Verhelst, Marian
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.01.2016
• Subjects: Application specific integrated circuits; Feature extraction; Heart rate variability; Light emitting diodes; Power demand; Silicon
• ISBN: 1467394661; 9781467394666
• ISSN: 2376-8606
• DOI: 10.1109/ISSCC.2016.7418069

Heart rate (HR) and its variability (HRV) provide critical information about an individual's cardiovascular and mental health state. In either application, long-term observation is crucial to arrive at conclusive decisions and provide useful diagnostic feedback [1]. Photoplethysmographic (PPG) estimation of HR and HRV has emerged as an attractive alternative to ECG, as it provides electrode-free operation increasing patient comfort. However, PPG monitoring systems robust to low ambient light conditions and low perfusion conditions require a LED as a light source, which strongly dominates the power consumption of the complete system. Compressive sampling (CS) based PPG readouts promise to mitigate this LED power consumption [2], yet require large computational power to recover the signal, hindering real-time embedded processing on energy-scarce wearable devices. This paper presents a fully integrated, low-power PPG readout ASIC, completely integrating a single-channel readout front-end (AFE) and a 12b SAR ADC and a digital back-end (DBE) for embedded energy-efficient real-time information extraction, that advances the state-of-the-art on the following fronts: 1) By smartly duty-cycling all system components synchronously on a sparse non-uniform CS sampling pulse stream, the LED driver power is reduced up to 30x, without significant loss of information. 2) Moreover, the necessity of wireless off-loading, or for computationally intensive embedded signal reconstruction, is circumvented by enabling the direct extraction of HR and HRV information from the compressed data in real-time on the ASIC, while consuming only 172μW for the complete system.