Estimating muscle activation from EMG using deep learning-based dynamical systems models

ObjectiveTo study the neural control of movement, it is often necessary to estimate how muscles are activated across a variety of behavioral conditions. However, estimating the latent command signal that underlies muscle activation is challenging due to its complex relation with recorded electromyographic (EMG) signals. Common approaches estimate muscle activation independently for each channel or require manual tuning of model hyperparameters to optimally preserve behaviorally-relevant features.ApproachHere, we adapted AutoLFADS, a large-scale, unsupervised deep learning approach originally designed to de-noise cortical spiking data, to estimate muscle activation from multi-muscle EMG signals. AutoLFADS uses recurrent neural networks (RNNs) to model the spatial and temporal regularities that underlie multi-muscle activation.Main ResultsWe first tested AutoLFADS on muscle activity from the rat hindlimb during locomotion, and found that it dynamically adjusts its frequency response characteristics across different phases of behavior. The model produced single-trial estimates of muscle activation that improved prediction of joint kinematics as compared to low-pass or Bayesian filtering. We also tested the generality of the approach by applying AutoLFADS to monkey forearm muscle activity from an isometric task. AutoLFADS uncovered previously uncharacterized high-frequency oscillations in the EMG that enhanced the correlation with measured force compared to low-pass or Bayesian filtering. The AutoLFADS-inferred estimates of muscle activation were also more closely correlated with simultaneously-recorded motor cortical activity than other tested approaches.SignificanceUltimately, this method leverages both dynamical systems modeling and artificial neural networks to provide estimates of muscle activation for multiple muscles that can be used for further studies of multi-muscle coordination and its control by upstream brain areas.

A Survey of Recent Indoor Localization Scenarios and Methodologies

Recently, various novel scenarios have been studied for indoor localization. The trilateration is known as a classic theoretical model of geometric-based indoor localization, with uniform RSSI data that can be transferred directly into distance ranges. Then, a trilateration solution can be algebraically acquired from theses ranges, in order to fix user’s actual location. However, the collected RSSI or other measurement data should be further processed and classified to lower the localization error rate, instead of using the raw data influenced by multi-path effect, multiple nonlinear interference and noises. In this survey, a large number of existing techniques are presented for different indoor network structures and channel conditions, divided as LOS (light-of-sight) and NLOS (non light-of-sight). Besides, the input measurement data such as RSSI (received signal strength indication), TDOA (time difference of arrival), DOA (distance of arrival), and RTT (round trip time) are studied towards different application scenarios. The key localization techniques like RSSI-based fingerprinting technique are presented using supervised machine learning methods, namely SVM (support vector machine), KNN (K nearest neighbors) and NN (neural network) methods, especially in an offline training phase. Other unsupervised methods as isolation forest, k-means, and expectation maximization methods are utilized to further improve the localization accuracy in online testing phase. For Bayesian filtering methods, apart from the basic linear Kalman filter (LKF) methods, nonlinear stochastic filters such as extended KF, cubature KF, unscented KF and particle filters are introduced. These nonlinear methods are more suitable for dynamic localization models. In addition to the localization accuracy, the other important performance features and evaluation aspects are presented in our paper: scalability, stability, reliability, and the complexity of proposed algorithms is compared in this survey. Our paper provides a comprehensive perspective to compare the existing techniques and related practical localization models, with the aim of improving localization accuracy and reducing the complexity of the system.

Unscented Kalman Filtering for Prognostics Under Varying Operational and Environmental Conditions

Prognostics gained a lot of research attention over the last decade, not the least due to the rise of data-driven prediction models. Also hybrid approaches are being developed that combine physics-based and data-driven models for better performance. However, limited attention is given to prognostics for varying operational and environmental conditions. In fact, varying operational and environmental conditions can significantly influence the remaining useful life of assets. A powerful hybrid tool for prognostics is Bayesian filtering, where a physical degradation model is updated based on realtime data. Although these types of filters are widely studied for prognostics, application for assets in varying conditions is rarely considered in literature. In this paper, it is proposed to apply an unscented Kalman filter for prognostics under varying operational conditions. Four scenarios are described in which a distinction is made between the level in which real-time and future loads are known and between short-term and long-term prognostics. The method is demonstrated on an artificial crack growth case study with frequently changing stress ranges in two different stress profiles. After this specific case, the generic application of the method is discussed. A positioning diagram is presented, indicating in which situations the proposed filter is useful and feasible. It is demonstrated that incorporation of physical knowledge can lead to highly accurate prognostics due to a degradation model in which uncertainty in model parameters is reduced. It is also demonstrated that in case of limited physical knowledge, data can compensate for missing physics to yield reasonable predictions.

SECEDO: SNV-based subclone detection using ultra-low coverage single-cell DNA sequencing

Recently developed single-cell DNA sequencing technologies enable whole-genome, amplifi-cation-free sequencing of thousands of cells at the cost of ultra-low coverage of the sequenced data(<0.05x per cell), which mostly limits their usage to the identification of copy number alterations(CNAs) in multi-megabase segments. Aside from CNA-based subclone detection, single-nucleotide vari-ant (SNV)-based subclone detection may contribute to a more comprehensive view on intra-tumorheterogeneity. Due to the low coverage of the data, the identification of SNVs is only possible whensuperimposing the sequenced genomes of hundreds of genetically similar cells. Here we present SingleCell Data Tumor Clusterer (SECEDO, lat. 'to separate'), a new method to cluster tumor cells basedsolely on SNVs, inferred on ultra-low coverage single-cell DNA sequencing data. The core aspects ofthe method are an efficient Bayesian filtering of relevant loci and the exploitation of read overlapsand phasing information. We applied SECEDO to a synthetic dataset simulating 7,250 cells and eighttumor subclones from a single patient, and were able to accurately reconstruct the clonal composition,detecting 92.11% of the somatic SNVs, with the smallest clusters representing only 6.9% of the totalpopulation. When applied to four real single-cell sequencing datasets from a breast cancer patient,SECEDO was able to recover the major clonal composition in each dataset at the original sequencingdepth of 0.03x per cell, an 8-fold improvement relative to the state of the art. Variant calling on theresulting clusters recovered more than twice as many SNVs with double the allelic ratio compared tocalling on all cells together, demonstrating the utility of SECEDO. SECEDO is implemented in C++ and is publicly available at https://github.com/ratschlab/secedo.

Deep state-space Gaussian processes

This paper is concerned with a state-space approach to deep Gaussian process (DGP) regression. We construct the DGP by hierarchically putting transformed Gaussian process (GP) priors on the length scales and magnitudes of the next level of Gaussian processes in the hierarchy. The idea of the state-space approach is to represent the DGP as a non-linear hierarchical system of linear stochastic differential equations (SDEs), where each SDE corresponds to a conditional GP. The DGP regression problem then becomes a state estimation problem, and we can estimate the state efficiently with sequential methods by using the Markov property of the state-space DGP. The computational complexity scales linearly with respect to the number of measurements. Based on this, we formulate state-space MAP as well as Bayesian filtering and smoothing solutions to the DGP regression problem. We demonstrate the performance of the proposed models and methods on synthetic non-stationary signals and apply the state-space DGP to detection of the gravitational waves from LIGO measurements.

State Estimation of an Underwater Markov Chain Maneuvering Target Using Intelligent Computing

In this study, an application of deep learning-based neural computing is proposed for efficient real-time state estimation of the Markov chain underwater maneuvering object. The designed intelligent strategy is exploiting the strength of nonlinear autoregressive with an exogenous input (NARX) network model, which has the capability for estimating the dynamics of the systems that follow the discrete-time Markov chain. Nonlinear Bayesian filtering techniques are often applied for underwater maneuvering state estimation applications by following state-space methodology. The robustness and precision of NARX neural network are efficiently investigated for accurate state prediction of the passive Markov chain highly maneuvering underwater target. A continuous coordinated turning trajectory of an underwater maneuvering object is modeled for analyzing the performance of the neural computing paradigm. State estimation modeling is developed in the context of bearings only tracking technology in which the efficiency of the NARX neural network is investigated for ideal and complex ocean environments. Real-time position and velocity of maneuvering object are computed for five different cases by varying standard deviations of white Gaussian measured noise. Sufficient Monte Carlo simulation results validate the competence of NARX neural computing over conventional generalized pseudo-Bayesian filtering algorithms like an interacting multiple model extended Kalman filter and an interacting multiple model unscented Kalman filter.