Arterial pulse wave has been considered as a vital sign in assessment of cardiovascular diseases. Noninvasive pulse sensor with compact structure, immunity to electro‐magnetic interference and high sensitivity is the research focus in recent years. While, optical fiber biosensor is a competitive option to meet these needs. Here, a diaphragm‐based optical fiber pulse sensor was proposed to achieve high‐precision radial pulse wave monitoring. A wearable device was developed, composed of a sports wristband and an aluminum diaphragm‐based optical fiber sensor tip of only 1 cm in diameter, which was highly sensitive to the weak acoustic signal. In particular, coherent phase detection was adopted to improve detection signal‐to‐noise ratio, so as to recover the high‐fidelity pulse waveforms. A clinical experiment was carried out to detect and morphological analyze the pulse waveforms of four subjects, the results of which preliminarily demonstrated the feasibility of pulse diagnosis method. The proposed pulse fiber sensor provides a comfortable way for pulse diagnosis, which is promising in early cardiovascular diseases indicating.

End-to-end approaches have drawn much attention recently for significantly simplifying the construction of an automatic speech recognition (ASR) system. RNN transducer (RNN-T) is one of the popular end-to-end methods. Previous studies have shown that RNN-T is difficult to train and a very complex training process is needed for a reasonable performance. In this paper, we explore RNN-T for a Chinese large vocabulary continuous speech recognition (LVCSR) task and aim to simplify the training process while maintaining performance. First, a new strategy of learning rate decay is proposed to accelerate the model convergence. Second, we find that adding convolutional layers at the beginning of the network and using ordered data can discard the pre-training process of the encoder without loss of performance. Besides, we design experiments to find a balance among the usage of GPU memory, training circle and model performance. Finally, we achieve 16.9% character error rate (CER) on our test set, which is 2% absolute improvement from a strong BLSTM CE system with language model trained on the same text corpus.

A smart mattress based on optical fiber Mach-zender interferometer (OF-MZI) is designed for noninvasive and continuous monitoring of human vital signs. Through arranging the sensing fiber between two elastic covering layers with sandwich structure, the mattress was sensitive to the respiration and heartbeat induced micro-pressure. In the processing terminal, the waveforms of vital signs were demodulated by 3 *3 coupler based differentiate and cross-multiplying method, and then four characteristic indicators including the heart rate, heartbeat amplitude, respiration rate, and respiration amplitude were respectively extracted through feature extraction algorithm, for evaluating the human health condition. Clinical experimental results of eighteen subjects indicate that the mattress system could not only distinguish the activity states of no body, on bed, body movement and off bed, but also contribute to clinical diagnosis of bradycardia, tachycardia, polypnea and apnoea. By adopting Bland-Altman analysis method, good reproducibility and accuracy were confirmed, where the max errors of heart rate and respiration rate are respectively 2 bpm and 1 bpm. Moreover, the responses at different positions of the mattress are identical and the continuous monitoring results in one day are consistent with daily change of vital signs, which proves that the fiber optic mattress has good reliability and stability. Beneficial from high-sensitivity, multiple parameters, long-term continuous monitoring, high comfortability and low cost, the mattress is promising in the early detection and prevention of cardiac and respiratory diseases as a household medical device.Index Terms-smart mattress, optical fiber Machzender interferometer, healthcare monitoring, heartbeat, respiration.

No abstract