Arshdeep Singh - Academia.edu (original) (raw)

Papers by Arshdeep Singh

2018 26th European Signal Processing Conference (EUSIPCO), 2018

Scene classification based on acoustic information is a challenging task due to various factors s... more Scene classification based on acoustic information is a challenging task due to various factors such as the nonstationary nature of the environment and multiple overlapping acoustic events. In this paper, we address the acoustic scene classification problem using SoundNet, a deep convolution neural network, pre-trained on raw audio signals. We propose a classification strategy by combining scores from each layer. This is based on the hypothesis that layers of the deep convolutional network learn complementary information and combining this layer-wise information provides better classification than the features extracted from an individual layer. In addition, we also propose a pooling strategy to reduce the dimensionality of features extracted from different layers of SoundNet. Our experiments on DCASE 2016 acoustic scene classification dataset reveals the effectiveness of this layer-wise ensemble approach. The proposed approach provides a relative improvement of approx. 30.85% over the classification accuracy provided by the best individual layer of SoundNet.

Materials Today: Proceedings, 2020

Wireless Sensor Networks are exposed to hostile and malignant environments, so they are vulnerabl... more Wireless Sensor Networks are exposed to hostile and malignant environments, so they are vulnerable to collusion attacks leading to compromised nodes and hence false data injection. Hence there is a requirement of secure data aggregation techniques. Iterative Filtering Algorithms provide secure data aggregation by imparting trustworthiness values to the nodes in the form of weight factors. In Iterative filtering algorithm, the most significant aspect for calculating weights is the selection of proper discriminant function. Different converging rates are furnished by different discriminant functions leading to different efficiencies. In this paper, we will be observing the effect of implementing the Iterative Filtering Algorithm with two new discriminant functions and comparing the convergence rates and hence the efficiency imparted by these two new discriminant functions with the existing discriminant functions. Also, we will be investigating and discussing the various material aspects for selecting suitable material for sensor nodes. We will also be considering the possibility of using biodegradable materials to manufacture sensor nodes.

Pattern Recognition Letters, 2020

Abstract In this letter, we propose a concise feature representation framework for acoustic scene... more Abstract In this letter, we propose a concise feature representation framework for acoustic scene classification by pruning embeddings obtained from SoundNet, a deep convolutional neural network. We demonstrate that the feature maps generated at various layers of SoundNet have redundancy. The proposed singular value decomposition based method reduces the redundancy while relying on the assumption that useful feature maps produced by different classes lie along different directions. This leads to ignoring the feature maps that produce similar embeddings for different classes. In the context of using an ensemble of classifiers on the various layers of SoundNet, pruning the redundant feature maps leads to reduction in dimensionality and computational complexity. Our experiments on acoustic scene classification demonstrate that ignoring 73% of feature maps reduces the performance by less than 1% with 12.67% reduction in computational complexity. In addition to this, we also show that the proposed pruning framework can be utilized to remove filters in the SoundNet network architecture, with 13x lesser model storage requirement. Also, the number of parameters reduce from 28 million to 2 million with marginal degradation in performance. This small model obtained after applying the proposed pruning procedure is evaluated on different acoustic scene classification datasets, and shows excellent generalization ability.

International Journal of Computer Applications, 2016

In this paper the effect on Zigbee mesh topology is analyzed by moving the nodes at different tra... more In this paper the effect on Zigbee mesh topology is analyzed by moving the nodes at different trajectories at different speed. The nodes are moved by using Helbert Space-filling curve, hexagon and outer square trajectory. The effect is analyzed in terms of load, delay and traffic received. Result shows that with change in trajectory the performance changes. Results have been analyzed by keeping 32 nodes fixed and all others moving at speed of 5 m/sec and 7 m/sec. It has been concluded that the hexagon trajectory performs better as compare to square trajectory at speed of 5 m/sec and at 7 m/sec when 32 nodes are kept fixed and all other are moving. Further it has been investigated that while moving 32 nodes and keeping all other fixed, the performance of square trajectory is better at speed of 5 m/sec and the performance of helbert curve is better at speed of 7 m/sec.

International Journal of Surgery Case Reports, 2011

INTRODUCTION: Isolated duplication of vas deferens is a rare anomaly with only eleven cases repor... more INTRODUCTION: Isolated duplication of vas deferens is a rare anomaly with only eleven cases reported in medical literature. Unawareness regarding this rare anomaly can lead to inadvertent injury to the vas during inguinal hernia surgery or failure of vasectomy procedure. PRESENTATION OF CASE: A 27-year-old gentleman was diagnosed with isolated duplication of vas during left sided open inguinal hernia surgery. He had no other genito-urinary anomaly. Patient had an uneventful recovery from surgery. CONCLUSION: It is a rare anomaly and unawareness regarding this condition can lead to catastrophic consequences during inguinal hernia and vasectomy surgeries.

2018 26th European Signal Processing Conference (EUSIPCO), 2018

Scene classification based on acoustic information is a challenging task due to various factors s... more Scene classification based on acoustic information is a challenging task due to various factors such as the nonstationary nature of the environment and multiple overlapping acoustic events. In this paper, we address the acoustic scene classification problem using SoundNet, a deep convolution neural network, pre-trained on raw audio signals. We propose a classification strategy by combining scores from each layer. This is based on the hypothesis that layers of the deep convolutional network learn complementary information and combining this layer-wise information provides better classification than the features extracted from an individual layer. In addition, we also propose a pooling strategy to reduce the dimensionality of features extracted from different layers of SoundNet. Our experiments on DCASE 2016 acoustic scene classification dataset reveals the effectiveness of this layer-wise ensemble approach. The proposed approach provides a relative improvement of approx. 30.85% over the classification accuracy provided by the best individual layer of SoundNet.

Materials Today: Proceedings, 2020

Wireless Sensor Networks are exposed to hostile and malignant environments, so they are vulnerabl... more Wireless Sensor Networks are exposed to hostile and malignant environments, so they are vulnerable to collusion attacks leading to compromised nodes and hence false data injection. Hence there is a requirement of secure data aggregation techniques. Iterative Filtering Algorithms provide secure data aggregation by imparting trustworthiness values to the nodes in the form of weight factors. In Iterative filtering algorithm, the most significant aspect for calculating weights is the selection of proper discriminant function. Different converging rates are furnished by different discriminant functions leading to different efficiencies. In this paper, we will be observing the effect of implementing the Iterative Filtering Algorithm with two new discriminant functions and comparing the convergence rates and hence the efficiency imparted by these two new discriminant functions with the existing discriminant functions. Also, we will be investigating and discussing the various material aspects for selecting suitable material for sensor nodes. We will also be considering the possibility of using biodegradable materials to manufacture sensor nodes.

Pattern Recognition Letters, 2020

Abstract In this letter, we propose a concise feature representation framework for acoustic scene... more Abstract In this letter, we propose a concise feature representation framework for acoustic scene classification by pruning embeddings obtained from SoundNet, a deep convolutional neural network. We demonstrate that the feature maps generated at various layers of SoundNet have redundancy. The proposed singular value decomposition based method reduces the redundancy while relying on the assumption that useful feature maps produced by different classes lie along different directions. This leads to ignoring the feature maps that produce similar embeddings for different classes. In the context of using an ensemble of classifiers on the various layers of SoundNet, pruning the redundant feature maps leads to reduction in dimensionality and computational complexity. Our experiments on acoustic scene classification demonstrate that ignoring 73% of feature maps reduces the performance by less than 1% with 12.67% reduction in computational complexity. In addition to this, we also show that the proposed pruning framework can be utilized to remove filters in the SoundNet network architecture, with 13x lesser model storage requirement. Also, the number of parameters reduce from 28 million to 2 million with marginal degradation in performance. This small model obtained after applying the proposed pruning procedure is evaluated on different acoustic scene classification datasets, and shows excellent generalization ability.

International Journal of Computer Applications, 2016

In this paper the effect on Zigbee mesh topology is analyzed by moving the nodes at different tra... more In this paper the effect on Zigbee mesh topology is analyzed by moving the nodes at different trajectories at different speed. The nodes are moved by using Helbert Space-filling curve, hexagon and outer square trajectory. The effect is analyzed in terms of load, delay and traffic received. Result shows that with change in trajectory the performance changes. Results have been analyzed by keeping 32 nodes fixed and all others moving at speed of 5 m/sec and 7 m/sec. It has been concluded that the hexagon trajectory performs better as compare to square trajectory at speed of 5 m/sec and at 7 m/sec when 32 nodes are kept fixed and all other are moving. Further it has been investigated that while moving 32 nodes and keeping all other fixed, the performance of square trajectory is better at speed of 5 m/sec and the performance of helbert curve is better at speed of 7 m/sec.

International Journal of Surgery Case Reports, 2011

INTRODUCTION: Isolated duplication of vas deferens is a rare anomaly with only eleven cases repor... more INTRODUCTION: Isolated duplication of vas deferens is a rare anomaly with only eleven cases reported in medical literature. Unawareness regarding this rare anomaly can lead to inadvertent injury to the vas during inguinal hernia surgery or failure of vasectomy procedure. PRESENTATION OF CASE: A 27-year-old gentleman was diagnosed with isolated duplication of vas during left sided open inguinal hernia surgery. He had no other genito-urinary anomaly. Patient had an uneventful recovery from surgery. CONCLUSION: It is a rare anomaly and unawareness regarding this condition can lead to catastrophic consequences during inguinal hernia and vasectomy surgeries.