An End-to-End Deep Neural Network for Autonomous Driving Designed for Embedded Automotive Platforms.

In this paper, one solution for an end-to-end deep neural network for autonomous driving is presented. The main objective of our work was to achieve autonomous driving with a light deep neural network suitable for deployment on embedded automotive platforms. There are several end-to-end deep neural networks used for autonomous driving, where the input to the machine learning algorithm are camera images and the output is the steering angle prediction, but those convolutional neural networks are significantly more complex than the network architecture we are proposing. The network architecture, computational complexity, and performance evaluation during autonomous driving using our network are compared with two other convolutional neural networks that we re-implemented with the aim to have an objective evaluation of the proposed network. The trained model of the proposed network is four times smaller than the PilotNet model and about 250 times smaller than AlexNet model. While complexity and size of the novel network are reduced in comparison to other models, which leads to lower latency and higher frame rate during inference, our network maintained the performance, achieving successful autonomous driving with similar efficiency compared to autonomous driving using two other models. Moreover, the proposed deep neural network downsized the needs for real-time inference hardware in terms of computational power, cost, and size.

Quantification of Finger-Tapping Angle Based on Wearable Sensors.

We propose a novel simple method for quantitative and qualitative finger-tapping assessment based on miniature inertial sensors (3D gyroscopes) placed on the thumb and index-finger. We propose a simplified description of the finger tapping by using a single angle, describing rotation around a dominant axis. The method was verified on twelve subjects, who performed various tapping tasks, mimicking impaired patterns. The obtained tapping angles were compared with results of a motion capture camera system, demonstrating excellent accuracy. The root-mean-square (RMS) error between the two sets of data is, on average, below 4°, and the intraclass correlation coefficient is, on average, greater than 0.972. Data obtained by the proposed method may be used together with scores from clinical tests to enable a better diagnostic. Along with hardware simplicity, this makes the proposed method a promising candidate for use in clinical practice. Furthermore, our definition of the tapping angle can be applied to all tapping assessment systems.

Stimulation map for control of functional grasp based on multi-channel EMG recordings.

Transcutaneous activation of muscles with electrical stimulation has limited selectivity in recruiting paralyzed muscles in stroke patients. However, the selectivity could be increased by the application of smaller electrodes and their appropriate positioning on the skin. We developed a method for selecting the appropriate positions of the stimulating electrodes based on electromyography (EMG). The EMG activity maps were estimated from signals recorded with two electrode arrays and two 24-channel wearable amplifiers positioned on the nonparetic and paretic forearms. The areas where the difference between the EMG maps obtained from the nonparetic and paretic arms was significant were identified as the stimulation sites. The stimulation was applied through array electrodes with magnetic holders and two wearable stimulators with four output channels each. The measures of functionality included joint angles measured with goniometers (hand opening) and grasp force measured with a multi-contact dynamometer (grasping). The stimulation protocol comprised co-activation of flexors and extensors to stabilize the wrist joint and prevent pronation/supination.

Quantitative and qualitative gait assessments in Parkinson's disease patients.

BACKGROUND/AIM: Postural impairments and gait disorders in Parkinson's disease (PD) affect limits of stability, impaire postural adjustment, and evoke poor responses to perturbation. In the later stage of the disease, some patients can suffer from episodic features such as freezing of gait (FOG). Objective gait assessment and monitoring progress of the disease can give clinicians and therapist important information about changes in gait pattern and potential gait deviations, in order to prevent concomitant falls. The aim of this study was to propose a method for identification of freezing episodes and gait disturbances in patients with PD. A wireless inertial sensor system can be used to provide follow-up of the treatment effects or progress of the disease. METHODS: The system is simple for mounting a subject, comfortable, simple for installing and recording, reliable and provides high-quality sensor data. A total of 12 patients were recorded and tested. Software calculates various gait parameters that could be estimated. User friendly visual tool provides information about changes in gait characteristics, either in a form of spectrogram or by observing spatiotemporal parameters. Based on these parameters, the algorithm performs classification of strides and identification of FOG types. RESULTS: The described stride classification was merged with an algorithm for stride reconstruction resulting in a useful graphical tool that allows clinicians to inspect and analyze subject's movements. CONCLUSION: The described gait assessment system can be used for detection and categorization of gait disturbances by applying rule-based classification based on stride length, stride time, and frequency of the shank segment movements. The method provides an valuable graphical interface which is easy to interpret and provides clinicians and therapists with valuable information regarding the temporal changes in gait.

Automatic identification and classification of freezing of gait episodes in Parkinson's disease patients.

Alternation of walking pattern decreases quality of life and may result in falls and injuries. Freezing of gait (FOG) in Parkinson's disease (PD) patients occurs occasionally and intermittently, appearing in a random, inexplicable manner. In order to detect typical disturbances during walking, we designed an expert system for automatic classification of various gait patterns. The proposed method is based on processing of data obtained from an inertial sensor mounted on shank. The algorithm separates normal from abnormal gait using Pearson's correlation and describes each stride by duration, shank displacement, and spectral components. A rule-based data processing classifies strides as normal, short (short(+)) or very short (short(-)) strides, FOG with tremor (FOG(+)) or FOG with complete motor block (FOG(-)). The algorithm also distinguishes between straight and turning strides. In 12 PD patients, FOG(+) and FOG(-) were identified correctly in 100% of strides, while normal strides were recognized in 95% of cases. Short(+) and short(-) strides were identified in about 84% and 78%. Turning strides were correctly identified in 88% of cases. The proposed method may be used as an expert system for detailed stride classification, providing warning for severe FOG episodes and near-fall situations.

Nonlinear optimization for drift removal in estimation of gait kinematics based on accelerometers.

A new data processing method is described for estimation of angles of leg segments, joint angles, and trajectories in the sagittal plane from data recorded by sensors units mounted at the lateral side of leg segments. Each sensor unit comprises a pair of three-dimensional accelerometers which send data wirelessly to a PC. The accelerometer signals comprise time-varying and temperature-dependent offset, which leads to drift and diverged signals after integration. The key features of the proposed method are to model the offset by a slowly varying function of time (a cubic spline polynomial) and evaluate the polynomial coefficients by nonlinear numerical simplex optimization with the goal to reduce the drift in processed signals (angles and movement displacements). The angles and trajectories estimated by our method were compared with angles measured by an optical motion capture system. The comparison shows that the errors for angles (rms) were below 4° and the errors in stride length were below 2%. The algorithm developed is applicable for real-time and off-line analysis of gait. The method does not need any adaptation with respect to gait velocity or individuality of gait.

Wireless distributed functional electrical stimulation system.

BACKGROUND: The control of movement in humans is hierarchical and distributed and uses feedback. An assistive system could be best integrated into the therapy of a human with a central nervous system lesion if the system is controlled in a similar manner. Here, we present a novel wireless architecture and routing protocol for a distributed functional electrical stimulation system that enables control of movement. METHODS: The new system comprises a set of miniature battery-powered devices with stimulating and sensing functionality mounted on the body of the subject. The devices communicate wirelessly with one coordinator device, which is connected to a host computer. The control algorithm runs on the computer in open- or closed-loop form. A prototype of the system was designed using commercial, off-the-shelf components. The propagation characteristics of electromagnetic waves and the distributed nature of the system were considered during the development of a two-hop routing protocol, which was implemented in the prototype's software. RESULTS: The outcomes of this research include a novel system architecture and routing protocol and a functional prototype based on commercial, off-the-shelf components. A proof-of-concept study was performed on a hemiplegic subject with paresis of the right arm. The subject was tasked with generating a fully functional palmar grasp (closing of the fingers). One node was used to provide this movement, while a second node controlled the activation of extensor muscles to eliminate undesired wrist flexion. The system was tested with the open- and closed-loop control algorithms. CONCLUSIONS: The system fulfilled technical and application requirements. The novel communication protocol enabled reliable real-time use of the system in both closed- and open-loop forms. The testing on a patient showed that the multi-node system could operate effectively to generate functional movement.

Kinematics of gait: new method for angle estimation based on accelerometers.

A new method for estimation of angles of leg segments and joints, which uses accelerometer arrays attached to body segments, is described. An array consists of two accelerometers mounted on a rigid rod. The absolute angle of each body segment was determined by band pass filtering of the differences between signals from parallel axes from two accelerometers mounted on the same rod. Joint angles were evaluated by subtracting absolute angles of the neighboring segments. This method eliminates the need for double integration as well as the drift typical for double integration. The efficiency of the algorithm is illustrated by experimental results involving healthy subjects who walked on a treadmill at various speeds, ranging between 0.15 m/s and 2.0 m/s. The validation was performed by comparing the estimated joint angles with the joint angles measured with flexible goniometers. The discrepancies were assessed by the differences between the two sets of data (obtained to be below 6 degrees) and by the Pearson correlation coefficient (greater than 0.97 for the knee angle and greater than 0.85 for the ankle angle).