Irina Hossain

An overview of heart-noise reduction of lung sound using wavelet transform based filter

Auscultation of the lung sound is a simple and noninvasive method to obtain some instant but useful informations to detect various respiratory diseases. However, one of the main problems in lung sound analysis is the interference of heart sounds, which is unavoidable during lung sound recording. Wavelet transformation based adaptive denoising technique has been proposed for heart sound reduction from lung sound. In the wavelet transform (WT) based filter it has been shown that the multiresolution representation of the lung sound signal in the WT domain combined with hard thresholding can separate the nonstationary part of the input signal (heart sound) from the stationary one (lung sound). This study investigated the spectral characteristics of the lung sound signals before and after WT based filtering. WT based filtering has been applied to lung sound signals recorded from anterior-right chest locations of six healthy subjects at low (7.5 ml/s/kg) and medium (15 ml/s/kg) flow rates. The R-waves of simultaneously recorded ECG signals were used to detect the lung sound segments at defined flow rates with inclusion or exclusion of heart sounds. The power spectra of the filtered lung sound segments were compared with the power spectra of the original sound segments, including and excluding heart sounds, at different frequency bands from 20-2400 Hz. Results show that the WT based filtering reduces the lung sound average power greatly over the whole frequency range. This results in pronounced change in the spectrum of the original signal that are of interest. Therefore, further investigations on feasibility of using this method for heart-noise reduction of lung sound signal are necessary.

Secure software upload in an intelligent vehicle via wireless communication links

The demand for drive-by-wire, telematics, entertainment, multimedia, pre-crash warning, highway guidance, remote diagnostic, etc. will significantly increase the complexity of a vehicles software modules. From time to time, the vehicles software may need to be updated due to many reasons such as the introduction of new features in vehicles, changing the navigation map, fixing software bugs, etc. Software updates must be done in secure modes to avoid any future disasters due to malfunctions of the vehicle. In this paper, we propose an architecture for secure software uploads in vehicles. We provide a detailed description of the secure software upload process.

Qualitative and quantitative evaluation of heart sound reduction from lung sound recordings

Recursive least squares (RLS) adaptive noise cancellation (ANC) and wavelet transform (WT) ANC have been applied and compared for heart sound (HS) reduction from lung sounds (LS) recordings. Novel processes for quantitative and qualitative evaluation of any method for HS reduction from LS have also been proposed.

Respiratory airflow estimation by acoustical means

The purpose of this study was to estimate respiratory airflow by acoustical means. Based on some preliminary results, a piecewise linear model was used to estimate respiratory airflow from lung sound average power. The model was tested using data from ten healthy subjects. Model coefficients were derived from a few breath sound segments with known airflow at medium flow rate and then applied to the rest of the lung sound segments to estimate airflow. The estimated airflow was compared with the actual recorded flow to calculate the error. The results showed an overall estimation error of 10.2 /spl plusmn/ 3.3%. The results also showed that the error at high (low) flow rate was mainly due to a consistent underestimation (overestimation) because the model coefficients were derived from medium flow rate. Therefore, it is suggested to reduce the error by applying a scaling factor to the proposed model.

Relationship between airflow and normal lung sounds

The relationship between airflow and respiratory sounds has always been of particular interest for diagnostic purposes. We investigated the relationship between airflow and lung sounds. Respiratory lung sound signals with the corresponding airflow signals from ten healthy subjects were studied. The lung sounds and the corresponding airflow signals were sequested to 100 ms segments with 50% overlap between successive segments. The mean lung sound amplitude (meanAMP) of the time signal as well as the average power (P/sub a/spl nu/g/) of the sounds between 100-300 Hz arid the corresponding mean airflow were calculated for each segment. The following five models were studied: linear relationship between meanAMP and flow (model 1), linear relationship between meanAMP and flow/sup 2/ (model 2), linear relationship between P/sub a/spl nu/g/ and flow (model 3), linear relationship between log(P/sub a/spl nu/g/) and flow (model 4) and linear relationship between log(P/sub a/spl nu/g/) and log(flow) (model 5). Linear regression analysis was used to investigate these relationships using the upper 15% of flow signal in each inspiration and the corresponding meanAMP and P/sub a/spl nu/g/. The correlation coefficient (r) between the model variables for each subject in each model was calculated. Also the RMS error between the regression line and the actual data for each model was averaged among the subjects. The results showed that the experimental data fits the model 3 very well. Therefore, we suggested a linear relationship between P/sub a/spl nu/g/ and flow: P/sub a/spl nu/g/=a/sub */flow+b. This model can be chosen to estimate airflow from the average power of lung sounds.

Finding the Lung sound-Flow Relationship in Normal and Asthmatic Subjects

To investigate the relationship between lung sound (LS) and flow, we studied LS signals from 5 healthy adults (group I), 10 healthy children (group II) and 7 asthmatic children (group III). The LS signals were recorded on right upper lung lobe at different flow rates varied from 0.4 to 3.0 L/s and the flow signals were measured at mouth. The LS and flow signals were parsed into segments of 1024 data points with 50% overlap between successive segments. The mean LS amplitude (mean AMP) and mean flow (flow) were calculated for each segment. The average power (Pave) of each segment was calculated from LS spectrum for different frequency bands between 20-600 Hz. Four different types of models, representing the relationship between mean AMP or Paveand flow, were investigated using different percentage of flow signal in each inspiratory phase. The model coefficients were derived from either linear regression analysis or polynomial curve fitting between the data and model variables. The correlation coefficients (r) between the experimental data and data estimated from the model coefficients were calculated for each subject in each model and averaged between the subjects. The results showed much stronger correlation between Paveand flow than mean AMP and flow for all groups. The best model to describe Paverelationship with flow was found to be power relationship in both healthy adults and children whereas a third-order polynomial curve best fitted the Paveand flow data in asthmatic group. The optimum frequency band to calculate Pavewas found to be 150-450 Hz for healthy subjects and 300-600 Hz for asthmatic children. The diminution of heart sound (HS) from LS recordings showed no change in the selected model in all three groups. The results of this study suggest the difference in Pave- flow relationship in healthy and asthmatic subjects may be used as a diagnostic tool for asthma.

Analysis of a Secure Software Upload Technique in Advanced Vehicles using Wireless Links

Software modules of an advanced vehicle can be updated using remote software upload (RSU) techniques. The RSU employs infrastructure-based wireless communication technique where the software supplier sends the software to the targeted vehicle via a roadside base station (BS). However, security is critically important in RSU to avoid any disasters due to malfunctions of the vehicle or to protect the proprietary algorithms from hackers, competitors or people with malicious intent. In this paper, we present a mechanism of secure software upload in an advanced vehicle. In order to increase the security level, we propose the vehicle to receive two copies of the software along with the message digest (MD) in each copy. The vehicle will install the new software only when it receives two identical copies of the software. To validate our proposition we find analytical expressions of average number of packet transmissions for successful software update. We investigate different cases depending on the vehicles buffer size and verification methods. Our analytical and simulation results show that it is sufficient to send two copies of software to the vehicle to thwart any security attack while uploading the software.

Analysis of Group Key Management Protocols for Secure Multicasting in Vehicular Software Distribution Network

In this paper we investigate the issues in designing key management architecture for secure multicasting in remote software distribution to advance vehicles. We consider the vehicular software distribution network (VSDN) as wireless network where vehicles are connected to the software distributors through roadside base stations (BSs). The large VSDN is divided into small regions and a Regional Group Manager (RGM) manages a region. Depending on the level of trust in the supporting BSs, we propose two group key management (GKM) protocols for the multicast groups (MGs) formed for software distribution. We categorize the VSDN as fully- trusted and semi-trusted systems. In a fully-trusted system the BSs have access to the multicast data and take part in GKM whereas in a semi-trusted system the BSs are not allowed to understand the multicast data, rather they act as proxies for the vehicles by honestly relaying the encrypted multicast data from the RGM to the vehicles. We compare the performance of these two protocols in terms of computation and communication overhead incurred on each entity involved during multicast session initialization, re-keying and handoff, storage overhead, security and scalability. Our analysis shows that the GKM protocol efficiency can be traded-off with the level of trust on the BSs.

Group Key Management for Secure Multicasting in Remote Software Upload to Future Vehicles

In future, updating various software modules in vehicles on a regular basis will be required for various reasons such as update functionalities in the existing system, add new functionalities, remove software bugs, update navigation map etc. For updating software to a large number of vehicles, remote updating using mobile multicasting would be the most efficient and economic than unicast updating in service station. However, the security requirement of multicast communication, i.e., confidentiality and integrity of the information transmitted and authenticity of the group members, is challenging. In this paper, we investigate issues in designing key management architectures for secure multicast network, particularly for remote software update in future vehicles. Vehicular software distribution network is considered as wireless network where vehicles are connected to the software distributors through base stations. Since the network consists of vehicles and base stations, the network dynamics is characterized by quasi-permanent mobility, high speed and frequent hand-off. High mobility and frequent hand-off increase the complexity of dynamic group membership which results in increase complexity in key generation and distribution. However, unlike the other mobile hosts such as PDA, vehicles provide sufficient battery, computational power and memory which allow performing complex cryptographic algorithm. Taking these things into consideration, we propose to use decentralized key management technique to generate and distribute multicast session key to the group members. Consequently, we evaluate our proposed architecture based on key distribution efficiency, re-keying efficiency, computational complexity, key storage requirement and scalability. From our analysis, it is found that the set-up and re-keying time of the proposed mechanism is in the range of few milliseconds while it provides confidentiality, authenticity and data integrity.

Performance Evaluation of Mobile Multicast Session Initialization Techniques for Remote Software Upload in Vehicle ECUs

Updating software in vehicle Electronic Control Units (ECUs) will become a mandatory requirement for a variety of reasons, for examples, to update functionality of an existing system, add new functionality, remove software bugs and to cope up with ITS infrastructure. Since multicasting is the most efficient method of group communication, updating software in an ECU of a large number of vehicles could benefit from it. In a previous work, we proposed an infrastructure-based mobile multicasting for remote software upload (RSU) in vehicle ECUs where an ECU receives the software from a remote software distribution center using the road side Base Stations (BSs) as gateways. Consequently, we proposed two multicast session establishment techniques based on the degree of trust on the BSs named Fully-trusted (FT) and Semi-trusted (ST) systems. In this paper, we conduct extensive simulation studies to evaluate the performance of these two techniques by finding the average latency to authenticate a vehicle and provide it with the multicast session key by the software provider during multicast session initialization. Simulation results show that the link establishment latency per vehicle of our proposed schemes is in the range of few seconds and the ST system requires two to three times higher time than the FT system.