Results and Discussions The experimental results show that the intensity of the interference peak in the transmission spectrum of the designed sandwich multimode fiber interferometer is very sensitive to micro-bending (Fig. 4). For sensors with different GIMMF lengths, the bending response of the characteristic peaks is different, but mainly shows the intensity change. When the GIMMF length is 2 mm within the curvature range of 0-2.36 m^-1, the characteristic peak intensity changes by 25 dB, with the most obviously changed overall intensity, largest linear region range, and maximum sensitivity up to 18.23 dB/m^-1. Therefore, a sandwich multimode fiber interferometer with the GIMMF length of 2 mm is selected for subsequent respiratory experiments. The study of respiratory signal noise reduction by low-pass filtering shows that this method can filter out most noise in respiratory signals (Fig. 6). Experimental studies for respiratory sensing indicate that the sensor can distinguish different types of respiratory conditions with universal applicability (Figs. 7-9). For steady-state respiratory signals with periodic regularity, it is accurate and effective to evaluate respiratory frequency by the dominant frequency in the fast Fourier transform (FFT) results of original respiratory signals (Fig. 7). For non-steady-state respiratory signals, the volunteers' respiratory rates can be displayed in real time using the STFT (Fig. 10). Finally, the performance comparison among the proposed sensor and other optical fiber respiratory sensors shows that the proposed sensor is characterized by extremely compact structure, high sensitivity, good stability, long service life, and anti-electromagnetic interference (Table 1).

Clinical data have demonstrated that respiratory rate (RR) is an important predictor of serious diseases including heart defects, heart failure, metabolic acidosis, and sleep apnea syndrome. Much important information related to physical conditions can be obtained by analyzing respiratory data. Flexible wearable devices can meet the needs of clinical medicine and health monitoring, which have attracted extensive attention. The most popular respiratory monitoring devices are based on electronic sensors, and cannot be employed in electromagnetic interference environments such as magnetic resonance imaging and computed tomography. In this regard, fiber optic sensors featuring high sensitivity, electromagnetic interference resistance, and corrosion resistance can overcome these challenges. Wearable respiratory sensing devices based on fiber sensors are mainly divided into curvature sensing and humidity sensing according to the principles. For the respiratory monitoring devices based on the humidity sensing principle, the optical fiber sensors have to be coated with moisture-sensitive materials, which have disadvantages such as time-consuming functionalization processes, uneven coating, and poor long-term stability in different degrees. In contrast, the respiratory monitoring system based on the curvature sensing principle is simpler and more stable. However, the compactness and sensitivity of the sensor still have great room for improvement. An optical fiber curvature sensor with ultra-high sensitivity and more compact size using two types of multimode fibers with mismatched core diameters is designed in this paper. Then, the proposed sandwich multimode fiber interferometer is integrated into an elastic waistband for respiratory sensing. The respiratory monitoring device is expected to be widely applied, with great potential in strong electromagnetic fields, radioactive examination environments (such as magnetic resonance imaging system and computed tomography), and sleep quality monitoring.

First, the proposed sandwich multimode fiber Mach-Zehnder interferometer is made by sandwiching the graded-index multimode fiber (GIMMF) between two pieces of very short stepped-index multimode fibers (SIMMFs) spliced with input-single mode fiber (SMF) and output-SMF, thus forming a SIMMF-GIMMF-SIMMF sensor structure. The core diameters of the SIMMFs and GIMMF are 105 μm and 50 μm respectively, and their cladding diameters are both 125 μm. Then, the effect of interference lengths on the curvature response of the SIMMF-GIMMF-SIMMF sensor is studied, and the optimal sensor parameters are selected according to the experimental results. After that, the designed interferometer is integrated into an elastic waistband with ultraviolet (UV) glue and fixed on the human abdomen. The respiratory signals of the volunteers are acquired in real time by monitoring the intensity changes of characteristic peaks in the transmission spectra of the sensor. The signals are denoised by low-pass filter, and the respiratory frequency is obtained by short-time Fourier transform (STFT). Finally, a series of respiratory sensing experiments (such as fast breathing, slow breathing, shallow breathing, and respiratory arrest) are conducted on multiple volunteers to verify the feasibility of the wearable respiratory sensor.

In this paper, a wearable respiratory sensor based on sandwich multimode fiber interferometer is proposed. The sensor unit is made by splicing a GIMMF with length of 1-3 mm between two SIMMFs with lengths of 1 mm. Due to the mismatching core diameters of GIMMF and SIMMF, the fiber Mach-Zehnder interference optical path is achieved. The interference peak intensity of the sensor is very sensitive to micro-bending, with a maximum sensitivity of -74.03 dB/m^-1 at the curvature range of 0-2.36 m^-1. Then, the sandwich multimode fiber optic interferometer is integrated into the elastic waistband and fixed on the human abdomen, and the respiratory signals can be obtained in real time accurately by monitoring the intensity change of the characteristic peaks in the transmission spectrum of the sensor. Experimental results show that the sensor can distinguish different respiratory conditions with universal applicability. The respiratory sensor is characterized by extremely compact structure, baseline drift without signals, high sensitivity, simple fabrication, low cost, easy integration, and electromagnetic interference resistance. It can be employed in strong electromagnetic fields or radioactive examination environments, such as magnetic resonance imaging systems and computed tomography.