Speech recognition has always been one of the research focuses in the field of human-computer communication and interaction. The main purpose of automatic speech recognition (ASR) is to convert speech waveform signals into text. Acoustic model is the main component of ASR, which is used to connect the observation features of speech signals with the speech modeling units. In recent years, deep learning has become the mainstream technology in the field of speech recognition. In this paper, a convolutional neural network architecture composed of VGG and Connectionist Temporal Classification (CTC) loss function was proposed for speech recognition acoustic model. Traditional acoustic model training is based on frame-level labels with cross-entropy criterion, which requires a tedious label alignment procedure. The CTC loss was adopted to automatically learn the alignments between speech frames and label sequences, such that the training process is end-to-end. The architecture can exploit temporal and spectral structures of speech signals simultaneously. Batch normalization (BN) technique was used for normalizing each layers input to reduce internal covariance shift. To prevent overfitting, dropout technique was used during training to improve network generalization ability. The speech signal was transformed into a spectral image through a series of processing to be the input of the neural network. The input feature is 200 dimensions, and output labels of acoustic mode is 415 Chinese pronunciation without pitch. The experimental results demonstrated that the proposed model achieves the Character error rate (CER) of 17.97% and 23.86% on public Mandarin speech corpus, AISHELL-1 and ST-CMDS-20170001_1, respectively.
We propose and experimentally demonstrate a low-cost, compact temperature-insensitive inclinometer, which is constructed by connecting a chemically etched fiber Bragg grating to a hollow-core fiber filled with tin. The optical power reflected from the grating is linearly proportional to the inclination angle of the grating and can provide a real-time measure of the inclination angle. Our experimental sensor can measure the inclination angle from 0° to 20° with an uncertainty of ± 0.35º and negligible temperature interference.
We propose and experimentally demonstrate a compact in-line micro-fiber inclinometer based on deformation of FBG, where the micro-fiber FBG beam is fabricated by special chemical etching method and the fiber pendulum is gained by splicing a section of hollow core fiber filling with tin to the end of the micro-fiber beam. The experiment results show that as the inclination angle increasing from00 to 200, the increments of the transmission loss and Bragg wavelength of the sensor are 1.81dB, and 0.035nm, respectively. Simultaneously, the change of the bandwidth at -25dBm increases linearly from 0.86nm to 1.048nm and the bandwidth sensitivity to inclination angle is 7.24pm/deg. On the other hand, temperature cross issue is solved by monitoring the bandwidth at -25dBm because the bandwidth sensitivity to temperature is 0.089pm/℃from 20℃ to 200℃.
An all-fiber accelerometer with temperature self-compensation based on Fabry-Pérot interferometer (FPI) and
simple-supported beam is proposed, where the two reflectors are the joint face between single mode fiber (SMF) and
hollow core fiber and the end face of the micro fiber used as vibration beam, respectively. The temperature effect could
be compensated by choosing proper material for the beam because it is movable in the FPI cavity. When the beam length
composed of SMF is 6 cm and the mass block is ~1.3218×10-7 Kg, the acceleration sensitivity is ~0.35 V/g and the
temperature sensitivity is less than 1pm/°C;.
A remote sensing system based on bandpass long-period fiber grating (LPFG) and fiber ring laser is proposed in this
paper. The reflective bandpass LPFG is fabricated by fusion splicing a piece of hollow core fiber (HCF) to the LPFG.
And the fiber ring laser is employed to narrow the line-width of the bandpass LPFG and realize remote sensing. As an
application example, a temperature experiment is conducted at the spot more than 1 km away from the fiber ring laser.
The experimental results show that such a sensor has a sensitivity of 0.0254nm/°C within the range of 20-150°C.