Arijit Chowdhury

Estimating true speed of moving vehicle using smartphone-based GPS measurement

The Global Positioning System (GPS) receivers are now an integral part of smartphones. However, phone based GPS measurements display much less accuracy as compared to professional grade receivers. On the other hand, the deep penetration of smartphones in consumer market offers opportunity for customizing new solutions. One such possible application is targeted towards identifying risky driving profile for the purpose of customizing auto-insurance premium. For this to be successful, one needs to estimate the true vehicle speed. In this paper, we have presented a method to estimate the true speed of a moving vehicle derived solely from GPS measurements. In this case the accelerometer sensors are not used in conjunction with GPS measurement. The results are compared with OBD2 speed measurement. The proposed method computes a better estimate of vehicle speed, where correctness is measured relative to OBD2 measurement.

An enhanced automated system for evaluating harsh driving using smartphone sensors

In this paper we propose an IoT based framework for driving style assessment using the driver owned smartphone as sensing platform. The GPS and the inertial sensors, embedded in the phone are used to continually sample measurements of position, speed and acceleration for the moving vehicle. The required computation is segregated between the phone and a remote server. We present the fusion algorithm which offers accurate estimation of velocity, thereby acceleration and position. To obviate the need for human intervention, an adaptive process in multi-sensor fusion is implemented. Since, the user is not expected to keep the phone in a predefined position and orientation (with respect to the vehicles longitudinal motion), an automated orientation correction module is essential and is described here.

Smartphone based estimation of relative risk propensity for inducing good driving behavior

Human activities analyses based on sensor data are gaining much importance. Of particular importance are those situations where man-machine interaction needs to be studied. The detection of risk induced while driving and customizing the insurance premium accordingly is an appropriate example. For such, the individuals are induced to utilize their personally owned smartphones in order to collect data and share them with insurance companies. The traditional approach counts the number of harsh events and infers the risk induced. However, such event-based inference is not necessarily a suitable approach for understanding each individuals propensity to indulge in high risk maneuvers. We propose an alternate method where a statistical route is adopted for quantifying risk propensity as well as a comparative analysis amongst a peer-group of drivers. A relatively moderate scale experimental test bed had been deployed by collecting driving related data (using their smartphones) from approximately 50 volunteers, fora duration of two months at a stretch. The experiment was unsupervised and the relative assessment is dependent on the quantity and quality of the collected data for each driver. In our model, the acceleration profiles displayed by each driver for every completed trip are observed to extract statistical features like Skewness squared and Kurtosis. It is observed that the kurtosis of the acceleration profiles stores major information about the driving styles. Subsequently, we have used statistical techniques to identify trends in data and used it to quantify the nature of the driving style. A comparative analysis within the peer-group (people with similar demographic features and similar work responsibilities) is done to judge individual propensities. It is envisaged that such application can be used to induce road safety through competitive spirit; additionally a large enterprise will find this tool useful to encourage employees to move towards a safe and fulfilling lifestyle. In this paper, we present the initial results of the above mentioned exercise using a smaller subset of the collected data.

Novel Statistical Post Processing to Improve Blood Pressure Estimation from Smartphone Photoplethysmogram

Blood pressure (BP) is considered to be an important biomarker for cardiac risk estimation. This paper deals with a non-conventional way of estimating BP using smartphone captured Photoplethysmogram (PPG) that enables unobtrusive health monitoring at home for possible alert generation. We have proposed a set of features that are independent to the inbuilt sensor of the capturing device. It is also observed that, BP estimated from a typical smartphone PPG signal fluctuates in successive cardiac cycles due to poor signal quality compared to a medical grade device. Hence, a novel post processing block is introduced, that rejects data depending on the BP distribution over all cardiac cycles in a session. Finally, Half Range Mode is used as a statistical average for the accepted sessions. This post processing methodology outperforms standard statistical averages in providing a better representative BP per session. The methodology yields mean absolute errors of 7.4% and 9.1% for predicting systolic and diastolic pressure respectively when validated over a dataset with a wide variation of BP.

An Improved Fusion Algorithm For Estimating Speed From Smartphone’s Ins/Gps Sensors

In recent times, number of researchers have investigated vehicle tracking applications by fusing the measurements done by accelerometers (as part of Inertial Navigation System-INS) and Global Positioning System (GPS). Since smartphones contain both the set of sensors, there exists a high degree of interest in utilizing personal phones for such tracking applications. However, mobile phone sensors have limitations in measurement accuracy and reliability. Usually, sudden changes in vehicle speed are not always captured well by GPS. Accelerometers, on the other hand, suffer from multiple noise sources. In this chapter, we investigate the noise performance of a few smartphone based accelerometers. Then, we apply the said noise analysis for improving the estimation of the speed of moving vehicle, as captured by GPS. A number of experiments were carried out to capture the vehicle’s position and speed from OBD2 (On Board Diagnosis V2), GPS as well as 3-axes accelerometer. We also demonstrate a method by which the phone’s orientation is compensated for while calculating speed from the measured acceleration. Further, a new method of INS/GPS fusion is proposed which enhances the accuracy of speed estimation. It is envisaged that with increasing estimation accuracy, the application of multi-sensor fusion in autonomous vehicles will be greatly enhanced.

A Novel Approach to the Identification of Compromised Pulmonary Systems in Smokers by Exploiting Tidal Breathing Patterns

Smoking causes unalterable physiological abnormalities in the pulmonary system. This is emerging as a serious threat worldwide. Unlike spirometry, tidal breathing does not require subjects to undergo forceful breathing maneuvers and is progressing as a new direction towards pulmonary health assessment. The aim of the paper is to evaluate whether tidal breathing signatures can indicate deteriorating adult lung condition in an otherwise healthy person. If successful, such a system can be used as a pre-screening tool for all people before some of them need to undergo a thorough clinical checkup. This work presents a novel systematic approach to identify compromised pulmonary systems in smokers from acquired tidal breathing patterns. Tidal breathing patterns are acquired during restful breathing of adult participants. Thereafter, physiological attributes are extracted from the acquired tidal breathing signals. Finally, a unique classification approach of locally weighted learning with ridge regression (LWL-ridge) is implemented, which handles the subjective variations in tidal breathing data without performing feature normalization. The LWL-ridge classifier recognized compromised pulmonary systems in smokers with an average classification accuracy of 86.17% along with a sensitivity of 80% and a specificity of 92%. The implemented approach outperformed other variants of LWL as well as other standard classifiers and generated comparable results when applied on an external cohort. This end-to-end automated system is suitable for pre-screening people routinely for early detection of lung ailments as a preventive measure in an infrastructure-agnostic way.

Non-destructive method to detect artificially ripened banana using hyperspectral sensing and RGB imaging

Fruits provide essential nutrition in most natural form suitable for human beings. They are best when ripened naturally. However, industrialization has provided many ways for quick ripening and for extended shelf life of fruits. Detection of artificial ripening could be done by sophisticated methods like chemical analysis in lab or visual inspection by experts, which may not be feasible all the time. Of all the fruits, banana is the most consumed fruit around the world. Adulteration of banana can have devastating effects on masses on scale. It is figured, bananas are potentially ripened using carcinogens like Calcium Carbide(CaC2). In this paper, we propose and devise a novel and automatic method to classify the naturally and artificially ripened banana using spectral and RGB data. Our results show that using a Deep Learning (Neural Network) on RGB data, we achieve accuracy of up-to 90%.and using Random Forest and Multilayer Perceptron (MLP) feed forward Neural Network as classifiers on spectral data we can achieve accuracies of up-to 98.74% and 89.49% respectively.

Investigations on Driver Unique Identification from Smartphone’s GPS Data Alone

Driver identification is an emerging area of interest in vehicle telematics, automobile control, and insurance. Recent body of works indicates that it may be possible to uniquely identify a driver using multiple dedicated sensors. In this paper, we present an approach for driver identification using smartphone GPS data alone. For our experiments, we collected data from 38 drivers for two months. We quantified the driver’s natural style by extracting a set of 137 statistical features from data generated for each completed trip. The analysis shows that, for the “driver identification” problem, an average accuracy of 82.3% is achieved for driver groups of 4-5 drivers. This is comparable to the state of the arts where mostly a multisensor approach has been taken. Further, it is shown that certain behavioral attributes like high driving skill impact identification accuracy. We observe that Random Forest classifier offers the best results. These results have great implications for various stakeholders since the proposed method can identify a driver based on his/her naturalistic driving style which is quantified in terms of statistical parameters extracted from only GPS data.