Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Stepping Motors: a Guide to Modern Theory and Practice

Pulmonary breathing sound plays a key role in the prevention and diagnosis of the lung diseases. Its correlation with pathology and physiology has become an important research topic in the pulmonary acoustics and the clinical medicine. However, it is difficult to fully describe lung sound information with the traditional features because lung sounds are complex and nonstationary signals. And the traditional convolutional neural network cannot also extract the temporal features of the lung sounds. To solve the problem, a lung sound recognition algorithm based on VGGish-BiGRU is proposed on the basis of transfer learning, which combines VGGish network with the bidirectional gated recurrent unit neural network (BiGRU). In the proposed algorithm, VGGish network is pretrained using audio set, and the parameters are transferred to VGGish network layer of the target network. The temporal features of the lung sounds are extracted through retraining BiGRU network with the lung sound data. During retraining BiGRU network, the parameters in VGGish layers are frozen, and the parameters of BiGRU network are fine-tuned. The experimental results show that the proposed algorithm effectively improves the recognition accuracy of the lung sounds in contrast with the state-of-the-art algorithms, especially the recognition accuracy of asthma.

/pdf/lung-sound-recognition-algorithm-based-on-vggish-bigru-3o2ql1t8q3.pdf

Lung Sound Recognition Algorithm Based on VGGish-BiGRU

In order to mitigate the power consumption issue for the sensor’s nodes, an efficient power optimized routing protocol is needed.
Therefore, this paper proposes a network coding backpressure routing (NCBPR) routing scheme for a large-scale Internet of Things (IoT) networks, which exploits the backpressure algorithm in order to divert packets flow. In the network, the packets are flowing from the highly congested node to low congested node, which helps to balance the load and optimized the fair use of the battery power of all the participating nodes. It divides the network into the small clusters, where the selection of clusters head depends upon an additional parameter of battery power apart from other optimum path parameters. It also employs an efficient data aggregation mechanism, which improves the throughput of the network by eliminating redundant packets. The network has been designed by considering 300 nodes in a network and the results have been drawn in terms of network throughput, packet delivery ratio and energy consumption. The results are presented in comparison with conventional well-known information-fusion-based role assignment (InFRA) and data routing for in-network aggregation (DRINA) routing schemes. The results prove that the proposed NCBPR scheme delivers significant improvement, such as throughput which increased by 21.38 and 12.13%, packet delivery ratio improved by 24.73 and 11.38%, and sensors node’s energy consumption is decreased by 61.46 and 44.35% as compared to conventional InFRA and DRINA schemes, respectively.

An optimal network coding based backpressure routing approach for massive IoT network

A reliable, remote, and continuous real-time respiratory sound monitor with automated respiratory sound analysis ability is urgently required in many clinical scenarios-such as in monitoring disease progression of coronavirus disease 2019-to replace conventional auscultation with a handheld stethoscope. However, a robust computerized respiratory sound analysis algorithm for breath phase detection and adventitious sound detection at the recording level has not yet been validated in practical applications. In this study, we developed a lung sound database (HF_Lung_V1) comprising 9,765 audio files of lung sounds (duration of 15 s each), 34,095 inhalation labels, 18,349 exhalation labels, 13,883 continuous adventitious sound (CAS) labels (comprising 8,457 wheeze labels, 686 stridor labels, and 4,740 rhonchus labels), and 15,606 discontinuous adventitious sound labels (all crackles). We conducted benchmark tests using long short-term memory (LSTM), gated recurrent unit (GRU), bidirectional LSTM (BiLSTM), bidirectional GRU (BiGRU), convolutional neural network (CNN)-LSTM, CNN-GRU, CNN-BiLSTM, and CNN-BiGRU models for breath phase detection and adventitious sound detection. We also conducted a performance comparison between the LSTM-based and GRU-based models, between unidirectional and bidirectional models, and between models with and without a CNN. The results revealed that these models exhibited adequate performance in lung sound analysis. The GRU-based models outperformed, in terms of F1 scores and areas under the receiver operating characteristic curves, the LSTM-based models in most of the defined tasks. Furthermore, all bidirectional models outperformed their unidirectional counterparts. Finally, the addition of a CNN improved the accuracy of lung sound analysis, especially in the CAS detection tasks.

/pdf/benchmarking-of-eight-recurrent-neural-network-variants-for-ji7bdjkdss.pdf

Benchmarking of eight recurrent neural network variants for breath phase and adventitious sound detection on a self-developed open-access lung sound database-HF_Lung_V1.

In this paper, a new algorithm of feature parameter extraction is proposed for application in speaker recognition system, which combines the traditional MFCC and the dynamic MFCC as a new series of coefficients. According to the statistics analysis of the different contribution by the dynamic MFCC and traditional MFCC, these coefficients are weighted as front-end parameters of the GMM, which would decrease the dimension of the mixed weighted GMM and reduce the computation complexity. The experiments based on the TIMIT and VOA speech database were implemented in MATLAB environment, and the results showed the speaker recognition system with the Weighted Dynamic MFCC could obtain better performance with high recognition rate and low computational complexity.

Speaker recognition using weighted dynamic MFCC based on GMM

In this paper, a new classification method is proposed to distinguish between normal and adventitious lung sounds. Firstly, using a dual channel electronical stethoscope, the active noise control (ANC) approach is employed to enhance the lung sounds which are subject to noise contamination in practical environment. Secondly, hidden Markov model (HMM) is utilized to characterize normal or adventitious lung sounds, which adopts maximum likelihood approach to calculate the acoustic likelihood for each respiratory phase. Thirdly, a deep neutral network (DNN) estimates the posterior probability of HMM for each observation, instead of the traditional Gaussian mixture model (GMM). And the acoustic input features of DNN consist of Mel frequency cepstral coefficients (MFCCs) extracted from each frame of lung sound samples. To evaluate the proposed DNN-HMM framework for lung sound classification, the recognition rate is compared between proposed method and traditional GMM-HMM approach. The results indicate that the proposed classification method can significantly improve the classification performance.

Classification between normal and adventitious lung sounds using deep neural network

Efficient architectures for realizing MDCT/IMDCT are presented. Based on the symmetry property of trigonometric functions, N-point MDCT formula will be transferred into an odd-even index paralleling process which can be achieved by two different types of decomposition, recursive formed DCT-II kernel or FFT-based DCT-IV kernel. Then a butterfly unit is employed to accelerate the computational speed. Furthermore, these new architectures are suitable for the computations of both MDCT and IMDCT, which improve the hardware efficiency. For verification, an experiment of MPEG2 AAC using this optimized FFT-based MDCT architecture on Altera Stratix FPGA is implemented. The analyzed results show the proposed structure provides a superior performance in terms of the computation rate, the data throughput and the hardware utilization. In addition, these MDCT/IMDCT architectures can be employed in many international audio standard systems.

Efficient architectures of MDCT/IMDCT implementation for MPEG audio codec

Recently x-vector has achieved a promising performance of speaker verification task and becomes one of the mainstream systems. In this paper, we analyzed the feature engineering based on the x-vector structure, and proposed a multi-feature integration method to further improve the feature representation of speaker characteristic. The proposed multi-feature integration method could be implemented in two ways, with the symmetric branches and the asymmetric branches, respectively, to incorporate different types of acoustic features in one neural network. While each branch processed one type of acoustic features on the frame level, the outputs of the two branches for each frame were spliced together as a super vector before being input into the statistics pooling layer. The experiments were executed on the VoxCeleb1 data set, and the results showed that the proposed multi-feature integration method obtained a 22.8% relative improvement over the baseline in EER value.

Speaker Embedding Extraction with Multi-feature Integration Structure

In this study, the authors consider the sensors' connection model design and energy optimisation of routing protocol in wireless sensor networks based on the Gaussian network connection model. Accordingly, by the node-symmetric and four different directions to four adjacent nodes of each node in the Gaussian network model, the study recommends a new wireless sensor network connection model, on which the network area will be divided into some virtual square grids. In the new wireless sensor network connection model, they describe each virtual square grid as a node in the Gaussian network, therefrom, they propose a routing method, which is a combination of the shortest path routing protocol in the Gaussian network and clustering protocol to improve the routing efficiency of the wireless sensor network. Some simulations were implemented in NS2, the results show that the proposed routing is very good.

https://ietresearch.onlinelibrary.wiley.com/doi/pdf/10.1049/iet-com.2018.5398

Lin Li

Papers

Speaker recognition using weighted dynamic MFCC based on GMM

Classification between normal and adventitious lung sounds using deep neural network

Efficient architectures of MDCT/IMDCT implementation for MPEG audio codec

Speaker Embedding Extraction with Multi-feature Integration Structure

Energy efficiency clustering based on Gaussian network for wireless sensor network