RRWaveNet: A Compact End-to-End Multiscale Residual CNN for Robust PPG Respiratory Rate Estimation

A Novel Respiratory Rate Estimation Algorithm from Photoplethysmogram Using Deep Learning Model

“…While increasing the window size tends to result in a better MAE, the significance of a smaller window size for rapid responsiveness to changes in RR cannot be understated. Despite RRWaveNet's higher MAE, optimal performance requires a larger window size, only achieving lower MAE values when using 32 and 64 s windows [24]. In this study, although the MAE is slightly higher than other deep learning algorithms, the window size is remarkably reduced to only 7 s, providing efficient RR estimation within a shorter timeframe for faster response.…”

“…Despite several studies employing deep learning algorithms, their performance has yet to surpass that of classical methods. Notably, current deep learning algorithms necessitate large window sizes for optimal performance, indicating a substantial gap in achieving superior results compared to classical methods [24]. Considering these observations, there is a compelling need to delve deeper into exploring deep learning algorithms for RR estimation, aiming to bridge the existing performance gap and potentially surpass the efficacy of classical methodologies.…”

“…In addition to that, ongoing research focuses on deep learning neural networks, with notable models like RRWaveNet ResNet, RespWatch, CycleGan and RespNet [24][25][26][27][28]. The state-of-the-art neural network is RRWaveNet, which has pushed the performance even further, achieving an MAE of 1.62 breaths/min and 1.59 breaths/min for BIDMC and Cap-noBase Datasets, respectively.…”

“…The state-of-the-art neural network is RRWaveNet, which has pushed the performance even further, achieving an MAE of 1.62 breaths/min and 1.59 breaths/min for BIDMC and Cap-noBase Datasets, respectively. However, it utilises a longer 32 and 64 s window size [24]. The RRWaveNet algorithm employs a sophisticated architecture featuring multiple convolutional neural networks (CNNs).…”

“…The RRWaveNet algorithm employs a sophisticated architecture featuring multiple convolutional neural networks (CNNs). Signal data for its training and evaluation are sourced from the BIDMC and CapnoBase datasets through oximeter recordings [24]. This algorithm demonstrates commendable performance in terms of MAE, showcasing its efficacy in accurately estimating respiratory rates.…”