• Publications
  • Influence
Speech Model Pre-training for End-to-End Spoken Language Understanding
A method to reduce the data requirements of end-to-end SLU in which the model is first pre-trained to predict words and phonemes, thus learning good features for SLU is proposed and improves performance both when the full dataset is used for training and when only a small subset is used.
Manifold regularized deep neural networks
A manifold learning based regularization framework for DNN training is presented to preserve the underlying low dimensional manifold based relationships amongst speech feature vectors as part of the optimization procedure for estimating network parameters.
Efficient keyword spotting using time delay neural networks
This paper describes a novel method of live keyword spotting using a two-stage time delay neural network trained using transfer learning and investigates various techniques to reduce computation in terms of multiplications per second of audio.
Tone Recognition Using Lifters and CTC
The proposed method for recognizing tones in continuous speech for tonal languages is shown to outperform the existing techniques in the literature in terms of tone error rate (TER).
Noise aware manifold learning for robust speech recognition
Noise aware manifold learning (NAML) is shown that NAML significantly reduces the speech recognition WER in a noisy speech recognition task over LPDA, particularly at low signal-to-noise ratios.
A Low Latency ASR-Free End to End Spoken Language Understanding System
This work proposes a speech understanding system with an additional constraint of designing a system that has a small enough footprint to run on small micro-controllers and embedded systems with minimal latency and a much smaller model when compared to other published works on the same task.
Multi-hop Multi-band Intelligent Relay-Based Architecture for LTE-Advanced Multi-hop Wireless Cellular Networks
A novel multi-hop multi-band intelligent (MMI) radio architecture is proposed for LTE-advanced cellular networks that would make use of a number of intelligent gateways in order to enable simultaneous usage of spectrum resources within the same cell.
Efficient manifold learning for speech recognition using locality sensitive hashing
This paper considers the application of a random projections based hashing scheme, known as locality sensitive hashing (LSH), for fast computation of neighborhood graphs in manifold learning based
On the development of variable length Teager energy operator (VTEO)
An alternative and generalized approach to TEO is suggested to calculate the instantaneous estimate of the energy where not only consecutive but other distant samples can also be incorporated in the calculation of running estimate ofThe energy and the number of samples taken to calculate energy can be increased depending on the signal properties.
A Family of Discriminative Manifold Learning Algorithms and Their Application to Speech Recognition
The proposed discriminative manifold learning approaches to feature space dimensionality reduction in noise robust automatic speech recognition are found to provide a significant reduction in word error rate with respect to the more well-known techniques for a variety of noise conditions.