In this paper, we provide a computational framework to automatically predict fluency in interface-based employment interviews. Fluency is known to influence the outcome of employment interviews. The interface-based interview setting is... more
In this paper, we provide a computational framework to automatically predict fluency in interface-based employment interviews. Fluency is known to influence the outcome of employment interviews. The interface-based interview setting is useful in assessing and giving feedback to the participants without any human intervention. To this end, we have collected a set of 106 interview videos from graduate students. Three external observers rate the interview videos for the variable of interest i.e., speaking fluency on a five point scale. We define several tasks based on grouping the fluency rating for easy prediction. We build a predictive model by first extracting linguistic and acoustic features automatically and then using machine learning algorithms like Linear Regression, Multi class Support Vector Machine (SVM) and Logistic Regression. We also analyze the role of different features and different categorizations towards characterization of speaking fluency.
Research Interests:
Because of their speed, versatility and accuracy, computers play an important role in medicine. They are useful in improving the efficiency and effectiveness of public health care systems. In hospitals they can be used for scheduling... more
Because of their speed, versatility and accuracy, computers play an important role in medicine. They are useful in improving the efficiency and effectiveness of public health care systems. In hospitals they can be used for scheduling admissions, for drug dosage control, patient monitoring in intensive care units and radiation therapy. They have been used with partial success in medical diagnosis. Computer aided instruction can very fruitfully supplement regular course work and hospital experience gained by medical students. In biomedical research, they have been used for studies of the structure of biomolecules, for biological signal processing, for contrast enhancement and improvement of resolution of X-ray photographs and scintiscan pictures and so on.In general, computers have been useful in improving the speed, efficiency and accuracy of existing methods, in enhancing the power of existing instruments and even making possible functions hitherto considered impossible.In India, th...
Research Interests:
There have been two major problems with use of computers by the 'uninitiated'; the need for keyboard skills and the restraint that communication has to be tightly structured, not like informal conversation as between two... more
There have been two major problems with use of computers by the 'uninitiated'; the need for keyboard skills and the restraint that communication has to be tightly structured, not like informal conversation as between two individuals. Using speech for man-machine communication obviates these problems and has been a long time dream for many. Research in speech recognition by computers has been on for over 40 years. Speech recognition has been an active area for research in India also for equally long. There is however, a long way still to go. This is because the problem is complex. For instance, an utterance might have significantly different properties even if it is spoken by the same speaker at different times – depending on context, mood and physical state of the speaker. Variability increases across speakers, depending on age, sex, region and cultural background. It is difficult for a speech recognition system to accommodate such wide variance. Also, speech is not produced...
The objective of the project was to develop an input/ output system to a computer with a facility for visual and voice feedback. Using primarily the speech mode, this was to be an interactive facility with provisions for keyboard entry,... more
The objective of the project was to develop an input/ output system to a computer with a facility for visual and voice feedback. Using primarily the speech mode, this was to be an interactive facility with provisions for keyboard entry, voice output and visual display. The aim was to make it possible for uninitiated users to interact with computers. Considering that there was no earlier work on speech in Indian languages a system working in a well defined and strictly delimited task environment was aimed at. It was to be speaker dependent with a vocabulary of about 200 words. The system was planned to accept clearly spoken isolated/ connected words and to produce intelligible speech. With the aim of realizing the above goals, several outputs were delivered at various stages of the project. The paper contains a description of major accomplishments with brief technical background. 5.1.2 OCR And Speech Recognition For Oriya Language Sanghamitra Mohanty, www.emille.lancs.ac.uk/ lesal/mo...
Research Interests:
A Voice Oriented Interactive Computing Environment (VOICE) has been implemented in the Hindi language. The system provides in interactive facility for visual and voice feedback. The 200 isolated word recognition system is designed around... more
A Voice Oriented Interactive Computing Environment (VOICE) has been implemented in the Hindi language. The system provides in interactive facility for visual and voice feedback. The 200 isolated word recognition system is designed around a railway reservation enquiry task and uses acoustic-phonetic segments as the basic units of recognition. Frame level classification into broad acoustic-phonetic categories is accomplished by a maximum likelihood classifier and segmentation by hierarchical clustering of the frame level likelihood vectors by use of explicit duration semi (Hidden) Markov Models. A more detailed classification of a few categories (vowels, voice bar and nasals in the first instance) is performed by neural nets. String matching using dynamic programming accomplishes lexical access, or conversion of the phonetic category symbol strings into words. Distributed processing of the word recognition task enables recognition at four times real time. A language processor disambiguates between multiple choices given by the recognizer for each word and even corrects some acoustic level recognition errors. This, the first system working in any Indian language, gives a recognition performance of 85% at the word level. For comparison, a purely HMM based word level recognizer has also been implemented. The performance is expected to improve further as there is still substantial scope for refinement.
Research Interests:
Research Interests:
Artificial Intelligence is an area of Computer Science concerned with making the computer perform tasks which, to be successfully done by human beings, require intelligence. Early efforts aimed at implementing general systems capable of... more
Artificial Intelligence is an area of Computer Science concerned with making the computer perform tasks which, to be successfully done by human beings, require intelligence. Early efforts aimed at implementing general systems capable of working in a wide variety of tasks as well as special systems doing only one type of task very well. Expert systems belong to the second category and aim at competence comparable to that of an expert, in a very well delimited area of activity. These consist of broadly two parts: a core consisting of the domain specific knowledge and inference rules and a shell which (is domain independent and) provides the facilities for using these and interacting with the user.The Fifth Generation Computer project proposed by Japan is the first comprehensive effort to consolidate and build up on the progress achieved in artificial intelligence and incorporate this into a new generation of very powerful computers, for use by the common man in his day to day life.Expert systems offer numer...
Research Interests:
... of the words or characters represented as long sequences of short line segments or at ... or ar-chetypal shapes are compared, training is dispensed with and robust writer independent ... A FEATURE-BASED APPROACH Statistical methods of... more
... of the words or characters represented as long sequences of short line segments or at ... or ar-chetypal shapes are compared, training is dispensed with and robust writer independent ... A FEATURE-BASED APPROACH Statistical methods of pattern recognition, in genera1 do not ...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Interfaces are the touch points between two or more entities which do not ‘speak the same language’ or work in ways that are dissimilar. A classical and extremely relevant example is the interaction between humans and machines. There is... more
Interfaces are the touch points between two or more entities which do not ‘speak the same language’ or work in ways that are dissimilar. A classical and extremely relevant example is the interaction between humans and machines. There is generally a gap between the two and an interfacing technology provides mechanisms that allow this gap to be bridged, thereby making it possible for these two different entities to communicate or talk to each other, Here we elaborate on how one can bridge the gap.
Research Interests:
Research Interests:
Research Interests:
... PVS Rao R. Raveendran Computer Systems and Communications Group, Tata Institute of Fundamental Research, Homi Bhabha Road, Bombay 400 005 ... The transition regions of isolated Consonant-Vowel (CV) utterances were manu-ally marked and... more
... PVS Rao R. Raveendran Computer Systems and Communications Group, Tata Institute of Fundamental Research, Homi Bhabha Road, Bombay 400 005 ... The transition regions of isolated Consonant-Vowel (CV) utterances were manu-ally marked and a data set of size 600 ...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
We describe a system for on line or off line recognition of cursive script. Our earlier work established that cursive script can be synthesised out of individual characters by using polynomial merging functions which satisfy boundary... more
We describe a system for on line or off line recognition of cursive script. Our earlier work established that cursive script can be synthesised out of individual characters by using polynomial merging functions which satisfy boundary conditions of continuity of the displacement functions x(t) and y(t) for each character and their first and second derivatives. We showed that even individual characters could be synthesised out of more primitive elements by using the same merging functions. The elements we choose are straight lines: not the usual line segments but a much smaller number of directed lines which we call shape vectors, ranging from only three vectors for simple characters such as e, 1 and o and a maximum of seven for m. We use slopes of the shape vectors and relative locations of points of maximum curvature (both highly quantised) as parameters for recognition. The system extracts parameters for individual characters from single specimens written in isolation and uses thes...
Research Interests:
Research Interests:
Research Interests:
We propose an efficient, self-organizing segmental measurement based on piecewise linear regression (PLR) fit of the short-term measurement trajectories. The advantages of this description are: (i) it serves to decouple temporal... more
We propose an efficient, self-organizing segmental measurement based on piecewise linear regression (PLR) fit of the short-term measurement trajectories. The advantages of this description are: (i) it serves to decouple temporal measurements from the recognition strategy; and, (ii) it leads to lesser computation as compared with conventional methods. Also, acoustic context can be easily integrated into this framework. The PLR
Research Interests:
Research Interests:
Research Interests:
To be associated with an important and challenging activity that is being carried out for the first time in the country is a truly memorable and satisfying, even if somewhat disquieting experience. I was privileged enough to have this... more
To be associated with an important and challenging activity that is being carried out for the first time in the country is a truly memorable and satisfying, even if somewhat disquieting experience. I was privileged enough to have this opportunity when I was involved in the design and implementation of India’s first computer. TIFRAC (Tata Institute of Fundamental Research Automatic Calculator) — so was it named by Jawaharlal Nehru when it was formally commissioned in 1960 — was a truly a unique computer. Some of the features of this machine are described later in this article. It is also my intention in this article to convey some of the excitement, adventure and sense of accomplishment that the successful completion of this exercise brought to the design team.
Research Interests:
... Neural Computation 2, 210215. Juang, BH and LR Rabiner (1985). Mixture autoregressive hidden Markov models for speech signals. IEEE Trans. Acoust. ... Juang, BH and Rabiner, LR, 1985. Mixtureautoregressive hidden Markov models for... more
... Neural Computation 2, 210215. Juang, BH and LR Rabiner (1985). Mixture autoregressive hidden Markov models for speech signals. IEEE Trans. Acoust. ... Juang, BH and Rabiner, LR, 1985. Mixtureautoregressive hidden Markov models for speech signals. IEEE Trans. Acoust. ...
Research Interests:
Research Interests:
Research Interests:
There are two possible approaches for dealing with user queries in an HMI system, namely. The simplistic keyword-based approach and the deep parsing approach. As we mentioned earlier neither of them are suitable for building a usable HMI... more
There are two possible approaches for dealing with user queries in an HMI system, namely. The simplistic keyword-based approach and the deep parsing approach. As we mentioned earlier neither of them are suitable for building a usable HMI system. In this chapter we suggest the use of a middle path, called the minimal parsing approach which is able to reliably identify the intent of the query and enable building a usable HMI system.
Research Interests:
This paper describes anAnd-or-Invert module developed for theOldap computer designed and being built at the Tata Institute of Fundamental Research. To obtain the maximum speed out of available transistors, the circuit makes use of... more
This paper describes anAnd-or-Invert module developed for theOldap computer designed and being built at the Tata Institute of Fundamental Research. To obtain the maximum speed out of available transistors, the circuit makes use of antisaturation and anti-cut-off techniques. The effect of different components on the transient response of the circuit is described. Detailed results of DC tolerance analysis and noise margins are included. The module which uses only indigenous components should be useful in any general digital system where speed is an important requirement.
Research Interests:
The speed of operation of double rank counters can be increased by a suitable modification of the gating logic now being used. The improvement in speed, predicted on theoretical grounds, has been experimentally verified. The prescribed... more
The speed of operation of double rank counters can be increased by a suitable modification of the gating logic now being used. The improvement in speed, predicted on theoretical grounds, has been experimentally verified. The prescribed logic enables the use of both the ranks of the counter to advantage, one rank counting in the normal, and the other in the reverse fashion.
Research Interests:
Research Interests:
In no field has the mutual reinforcement of the development of device technology and system applications been more pronounced, wide spread or faster than in the area of computers. The rapid improvements in raw computational power... more
In no field has the mutual reinforcement of the development of device technology and system applications been more pronounced, wide spread or faster than in the area of computers. The rapid improvements in raw computational power available, cost-performance ratio, reliability, size and power consumption have all been remarkable in the recent past. Initially, computers could be visualized only as isolated stand-alone systems. Recent trends in minicomputers and microcomputers have however permitted them to be used as subsystems in larger systems and even as components in small subsystems and units.The availability of raw computational power and high information handling capabilities make large computers useful in such diverse applications as numerical weather prediction, management information systems and iiuclear research. The high reliability, speed and versatility of minicomputers are responsible for their being used increasingly in route switching and message switching applications in communications, on...
Research Interests:
In this paper, we address the problem of synthesizing connected handwritten script from individual characters written in isolation. Connected writing is viewed as a natural evolution from writing t...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
The Burg maximum-entropy (ME) method of spectral analysis is used to estimate the short-time spectral envelope of the voiced speech-signal. A number of vowel segments from continuous speech are analysed using this method. The Burg ME... more
The Burg maximum-entropy (ME) method of spectral analysis is used to estimate the short-time spectral envelope of the voiced speech-signal. A number of vowel segments from continuous speech are analysed using this method. The Burg ME method is compared with the autocorrelation and covariance methods of linear prediction using normalized linear prediction error and accuracy in estimating the speech spectrum as criteria. The results obtained from pitch-synchronous and pitch-asynchronous analysis of vowel segments are discussed.