Mobile Computing

Mobile phones have emerged as truly pervasive and affordable Information and Communication Technology (ICT) platform in the last decade. Large penetration of cellular networks and availability of advanced hardware platforms have inspired multiple innovative research opportunities in mobile computing domain. However, most of the research challenges have focused on typical scenarios existing in the developed economies. Using the healthcare as an application for mobile computing, we present research challenges and novelties in mobile computing domain that take account for differences between developing such as India and developed economies based on commonly available mobile platforms, communication cost, differences in user behavior and acceptable societal norms, among others. With localization being a critical component of various mobile systems, we present early results for the application of GPS-less localization that highlight the corresponding differences in Indian context.

Location is a primary indicator of context and forms the core basis of several context-aware applications. Most common way of getting location information is to use specialized hardware like GPS. However, GPS is expensive and is available only on high-end phones restricting its use to a smaller population in developing countries. Further, GPS also consumes a lot of battery power during its operation, thereby making it infeasible to run for longer durations with limited mobile phone battery. An alternative to GPS-based localization is GSM-based localization that is more suitable for developing countries due to much lower power consumption and ability to run even on low-end phones. Currently available, network-operator independent, GSM-based solutions require building perceptual map of cell towers in a city using war-driving.

Due to increase in use of Short Message Service (SMS) over mobile phones in developing countries, there has been a burst of spam SMSes. Content-based machine learning approaches were effective in filtering email spams. Researchers have used topical and stylistic features of the SMS to classify spam and ham. SMS spam filtering can be largely influenced by the presence of regional words, abbreviations and idioms. We have tested the feasibility of applying Bayesian learning and Support Vector Machine(SVM) based machine learning techniques which were reported to be most effective in email spam filtering on a India centric dataset. In our ongoing research, as an exploratory step, we have developed a mobile-based system SMSAssassin that can filter SMS spam messages based on bayesian learning and sender blacklisting mechanism. Since the spam SMS keywords and patterns keep on changing, SMSAssassin uses crowd sourcing to keep itself updated. Using a dataset that we are collecting from users in the real-world, we evaluated our approaches and found some interesting results.

Location-aware applications are steadily gaining popularity across the world. However, lack of GPS in low-end programmable phones (< $100) and absence of Wi-Fi infrastructure in developing countries prevent users of low-end phones (majority of population in developing countries) from using such applications as their phones can not get location. GSM-based approaches, such as using Cell IDs, have been developed as they do not require specific hardware on phone and need no additional infrastructural support. However, Cell ID based approaches require access to a comprehensive database of Cell IDs. Such a database does not exist or is very limited in developing countries and its growth is not promising either.

In developing countries, smartphones and high bandwidth connections are not yet within the reach of majority of the people. To share multimedia content among themselves, people rely on manual local sharing mechanisms i.e. either using physical exchange of memory card or through short-range wireless technologies, such as Bluetooth. However, this model of content sharing is not scalable in terms of number of users and shared items.

Large proliferation of mobile phone applications result in extensive use of data intensive services such as multimedia download and social network communication. With limited penetration of 3G/4G networks in developing countries, it is common to use low bandwidth 2G services for data communication, resulting in larger download time and correspondingly high power consumption.

With the growing number of cities and population, continuous monitoring of city's infrastructure and automated collection of day-to-day events (such as traffic jam) is essential and can help in improving life style of citizens. It is extremely costly and ineffective to install hardware sensors to sense these events in developing regions. Due to advent of smartphones, citizens can play role of sensors and actively participate in collection of the events which can be shared with others for information or can be used in decisions which affects city development.

Short Message Service (SMS) is one of the most frequently used services in the mobile phones, next to calls. In developing countries like India, SMS is the cheapest mode of communication. The advantage of this fact is exploited by the advertising companies to reach masses. The unsolicited SMS messages (a.k.a. spam SMS) generates notifications, thus consuming precious user attention. To formulate spam SMS problem and understand user's needs and preceptions, we conducted an online survey with 458 participants in different cities of India. Most of the survey participants admitted that they are quite annoyed with burst of SMS spams and ineffectiveness of regulatory solutions. However, some participants reported that, they do get useful information from spam SMSes sometime(e.g. discounts at a popular food joint). In this paper, we present design and implementation of a usercentric spam SMS filtering application i.e. SMSAssassin that uses content based machine learning techniques with user generated features to filter unwanted SMSes and reduces the burden of notifications for a mobile user.

Methods based on combinatorial graph cut algorithms received a lot of attention in the recent years for their robustness as well as reasonable computational demands. These methods are built upon an underlying Maximum a Posteriori estimation of Markov Random Fields and are suitable to solve accurately many different problems in image analysis, including image segmentation. In this paper we present a twostage graph cut based model for segmentation of touching cell nuclei in fluorescence microscopy images. In the first stage voxels with very high probability of being foreground or background are found and separated by a boundary with a minimal geodesic length. In the second stage the obtained clusters are split into isolated cells by combining image gradient information and incorporated a priori knowledge about the shape of the nuclei. Moreover, these two qualities can be easily balanced using a single user parameter. Preliminary tests on real data show promising results of the method.