Amitrajit Bhattacharjee

Amitrajit Bhattacharjee

Auto-generated profile, based on publicly available information

Amitrajit Bhattacharjee

School of Interactive Computing, Georgia Institute of Technology, Atlanta, Georgia, USA

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 August 2024 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
ZLZikang Leng
Zikang Leng
ABAmitrajit Bhattacharjee
Amitrajit Bhattacharjee
HRHrudhai Rajasekhar
Hrudhai Rajasekhar
LZLizhe Zhang
Lizhe Zhang
EBElizabeth Bruda
Elizabeth Bruda
TPThomas Ploetz
Thomas Ploetz
One of the primary challenges in the field of human activity recognition (HAR) is the lack of large labeled datasets. This hinders the development of robust and generalizable models. Recently, cross modality transfer approaches have been explored that can alleviate the problem of data scarcity. These approaches convert existing datasets from a source modality, such as video, to a target modality, such as inertial measurement units (IMUs). With the emergence of generative AI models such as large language models (LLMs) and text-driven motion synthesis models, language has become a promising source data modality as well - as shown in proof of concepts such as IMUGPT. In this work, we conduct a large-scale evaluation of language-based cross modality transfer to determine their effectiveness for HAR. Based on this study, we introduce two new extensions for IMUGPT that enhance its use for practical HAR application scenarios: a motion filter capable of filtering out irrelevant motion sequences to ensure the relevance of the generated virtual IMU data, and a set of metrics that measure the diversity of the generated data facilitating the determination of when to stop generating virtual IMU data for both effective and efficient processing. We demonstrate that our diversity metrics can reduce the effort needed for the generation of virtual IMU data by at least 50%, which opens up IMUGPT for practical use cases beyond a mere proof of concept.

20 September 2023 in Transactions of the International Society for Music Information Retrieval
Verified
Journal Article
Open access
PDF file
RARohit M. Ananthanarayana
Rohit M. Ananthanarayana
ABAmitrajit Bhattacharjee
Amitrajit Bhattacharjee
PRPreeti Rao
Preeti Rao
Motivated by the musicological relevance of tabla stroke categories in tabla accompaniment playing, we present an automatic four-way stroke classification system based on convolutional neural networks, while recognising the challenge of instrument- and style-independent classification with limited available labeled training data. Tabla stroke transcription has been traditionally viewed as a monophonic timbre recognition task given the variety of musically distinct single-drum and two-drum strokes that comprise the music. In this work, we adopt a more sound-production based approach by identifying a reduced set of ‘atomic’ strokes (damped, resonant treble and resonant bass) that serve as the primary level for classification. An advantage of this is the better exploitation of tabla training data and the potential for better generalization. The new viewpoint also facilitates exploring the acoustic similarity with Western drums via the investigation of transfer learning for the tabla task. We find that the drum pretraining learns features that are useful for our tabla stroke classification task. Further fine-tuning the model with the target tabla data leads to the expected improvements in performance, which, however, surpasses that achieved with a purely tabla-trained model for only one of the stroke categories.