Browsing by Author "Rajan, M.R."

Now showing 1 - 3 of 3

A Hardware Accelerator Based on Quantized Weights for Deep Neural Networks
(Springer Verlag service@springer.de, 2019) Sreehari, R.; Deepu, V.; Rajan, M.R.
The paper describes the implementation of systolic array-based hardware accelerator for multilayer perceptrons (MLP) on FPGA. Full precision hardware implementation of neural network increases resource utilization. Therefore, it is difficult to fit large neural networks on FPGA. Moreover, these implementations have high power consumption. Neural networks are implemented with numerous multiply and accumulate (MAC) units. The multipliers in these MAC units are expensive in terms of power. Algorithms have been proposed which quantize the weights and eliminate the need of multipliers in a neural network without compromising much on classification accuracy. The algorithms replace MAC units with simple accumulators. Quantized weights minimize the weight storage requirements. Quantizing inputs and constraining activations along with weights simplify the adder as well as further reduce the resource utilization. A systolic array-based architecture of neural network has been implemented on FPGA. The architecture has been modified according to Binary Connect and Ternary Connect algorithms which quantize the weights into two and three levels, respectively. The final variant of the architecture has been designed and implemented with quantized inputs, Ternary connect algorithm and activations constrained to +1 and âˆ’1. All the implementations have been verified with MNIST data set. Classification accuracy of hardware implementations has been found comparable with its software counterparts. The designed hardware accelerator has achieved reduction in flip-flop utilization by 7.5 times compared to the basic hardware implementation of neural network with high precision weights, inputs and normal MAC units. The power consumption also has got reduced by half and the delay of critical path decreased by three times. Thus, larger neural networks can be implemented on FPGA that can run at high frequencies with less power. Â© 2019, Springer Nature Singapore Pte Ltd.
Predicting Gamakas-The Essential Embellishments in Karnatic Music
(2019) Rajan, M.R.; Vijayasenan, D.; Vijayakumar, A.
Gamakas are the musical embellishments used in Karnatic Music. Predicting them from the musical notations plays an important part in applications like automatic synthesis and composition of Karnatic Music. Since there are no well-defined rules governing the use of gamakas, predicting them is a challenging problem. In this work, we propose a method to detect the presence and type of gamakas, in a data-driven manner, from the annotated symbolic music alone. We propose features based on the notes of the song for these tasks. These features are used as inputs to a Random Forest Classifier. We digitise 80 songs from a well known reference book of Karnatic music to create a dataset consisting roughly 30000 notes. We train the classifier on around 12000 notes and test on roughly 18000 notes. From our experiments, the accuracy values obtained for predicting gamaka presence and type are 77% and 70%, respectively. These are significantly better than random classification accuracies. We also analyse the importance of neighbourhood of notes for the detection and classification of gamakas. It is observed that the best accuracy is obtained for gamaka presence detection when a both-sided neighbourhood of size three is considered; and best accuracy for gamaka type prediction is obtained with a both-sided neighbourhood of size one. The analysis performed on the training data reveals that there is information contained in these neighbourhoods for distinguishing between gamaka and non-gamaka notes. 2013 IEEE.
Predicting Gamakas-The Essential Embellishments in Karnatic Music
(Institute of Electrical and Electronics Engineers Inc., 2019) Rajan, M.R.; Vijayasenan, D.; Vijayakumar, A.
Gamakas are the musical embellishments used in Karnatic Music. Predicting them from the musical notations plays an important part in applications like automatic synthesis and composition of Karnatic Music. Since there are no well-defined rules governing the use of gamakas, predicting them is a challenging problem. In this work, we propose a method to detect the presence and type of gamakas, in a data-driven manner, from the annotated symbolic music alone. We propose features based on the notes of the song for these tasks. These features are used as inputs to a Random Forest Classifier. We digitise 80 songs from a well known reference book of Karnatic music to create a dataset consisting roughly 30000 notes. We train the classifier on around 12000 notes and test on roughly 18000 notes. From our experiments, the accuracy values obtained for predicting gamaka presence and type are 77% and 70%, respectively. These are significantly better than random classification accuracies. We also analyse the importance of neighbourhood of notes for the detection and classification of gamakas. It is observed that the best accuracy is obtained for gamaka presence detection when a both-sided neighbourhood of size three is considered; and best accuracy for gamaka type prediction is obtained with a both-sided neighbourhood of size one. The analysis performed on the training data reveals that there is information contained in these neighbourhoods for distinguishing between gamaka and non-gamaka notes. © 2013 IEEE.