Nandini Bondale

A feature-based hierarchical speech recognition system for Hindi

This paper presents a description of a speech recognition system forHindi. The system follows a hierarchic approach to speech recognition and integrates multiple knowledge sources within statistical pattern recognition paradigms at various stages of signal decoding. Rather than make hard decisions at the level of each processing unit, relative confidence scores of individual units are propagated to higher levels. Phoneme recognition is achieved in two stages: broad acoustic classification of a frame is followed by fine acoustic classification. A semi-Markov model processes the frame level outputs of a broad acoustic maximum likelihood classifier to yield a sequence of segments with broad acoustic labels. The phonemic identities of selected classes of segments are decoded by class-dependent neural nets which are trained with class-specific feature vectors as input. Lexical access is achieved by string matching using a dynamic programming technique. A novel language processor disambiguates between multiple choices given by the acoustic recognizer to recognize the spoken sentence.

mHEALTHPHC: An ICT Tool for Primary Healthcare in India

Access to timely, quality healthcare is one of the important issues concerning communities in rural India, among many others such as education, safe drinking water, transport, approach roads, and electricity. Information and Communications Technology (ICT) has made a significant impact on the functioning of enterprises and urban communities; however, the benefits of ICT are yet to reach rural communities. This article describes how an innovative ICT tool called “mHEALTHPrimary Health Care” (mHEALTH-PHC) can be used to meet the primary healthcare requirements of the community. The MHEALTH-PHC tool was developed after studying the needs and the logistic problems related to primary healthcare of the rural tribal population of villages in India.

Distributed processing and internet technology to solve challenges of primary healthcare in india

National Reproductive and Child Health program, focused on Mother and Child Health, part of Millennium Developmental Goals, is being implemented through Primary Health Centers of the Public Healthcare System. The health record generated through community health interventions are manual, leading to undue delay in diagnosis and emergency care. In this paper, we show how Distributed Processing and Internet Technology can be applied through innovative platform called mHEALTH-PHC, to provide timely, quality healthcare to remote population using existing infrastructure. mHEALTH-PHC combines client server, cellular, mobile phone technologies and medical test equipment to establish two way connection between patient in a village and Public Healthcare System. Our field study shows that mHEALTH-PHC can be effective in health surveillance, thereby leading to prompt, efficient, quality healthcare. We take the consortium approach involving IT and Public Health experts, Directorate of Health Services, pharmaceutical and health insurance industries, to make quality healthcare affordable and sustainable.

Prosodic features of Marathi news reading style

Text-to-speech synthesizers present an attractive alternative to reading in hands-free communication scenarios. Speech intelligibility and naturalness are key to the user acceptability of synthesized speech. The accurate modeling of prosody plays an important role in both dimensions. While prosody is language dependent, it is also strongly dependent on the speaking style. In this work, we study the important prosodic features of news reading style in Marathi using publicly available radio broadcasts. Prominence and boundaries are among the important linguistic cues conveyed via a news readers prosody. Using perception testing, we obtain boundaries and prominent words in broadcast recordings of two female news readers. We measure acoustic parameters known to serve as cues to prominence such as the fundamental frequency, duration and intensity. We also make observations on timing and pitch phenomena at inter- and intra-sentence breaks. Our results indicate that prominence depends strongly on achieved FO span in the word and to a smaller extent on duration increase. Breaks are signaled by pauses and pre-boundary lengthening of the final syllable. We observe that, unlike English, sentence ending in Marathi is not always accompanied by a pitch fall in the final syllable. The implications of these observations on prosody generation are discussed.

Multi-speaker, narrowband, continuous Marathi speech database

We describe the development of a continuous speech database in Marathi language. Speech data was collected from about 1500 literate speakers from 34 districts of Maharashtra, with a variety of characteristics such as age group, gender, mother tongue and educational qualification. The subjects called the data acquisition system with personal mobile handsets, and read specially designed sentence sets. The sentence data acquisition process was conducted on field in contrast to a quiet environment. As a result, the acquired speech data captured large amount of nonspeech sounds as well as incompletely spoken words. So, the speech data was transcribed employing additional labels to denote frequently occurring nonspeech sounds, different kinds of incomplete words and invalid words. We characterize the database in terms of the statistics of features such as gender distribution of speakers, phonemic richness, amount of non speech sounds, and average sentence and word lengths for both reference and actual sentences.