Syed Muslim Shah

The authors Syed Muslim Shah and Syed Asghar Ali Shah were not included as authors in the original publication

Automatic Speech Recognition, (ASR) has achieved the best results for English, with end-to-end neural network based supervised models. These supervised models need huge amounts of labeled speech data for good generalization, which can be quite a challenge to obtain for low-resource languages like Urdu. Most models proposed for Urdu ASR are based on Hidden Markov Models (HMMs). This paper proposes an end-to-end neural network model, for Urdu ASR, regularized with dropout, ensemble averaging and Maxout units. Dropout and ensembles are averaging techniques over multiple neural network models while Maxout are units in a neural network which adapt their activation functions. Due to limited labeled data, Semi Supervised Learning (SSL) techniques are also incorporated to improve model generalization. Speech features are transformed into a lower dimensional manifold using an unsupervised dimensionality-reduction technique called Locally Linear Embedding (LLE). Transformed data along with higher dimensional features is used to train neural networks. The proposed model also utilizes label propagation-based self-training of initially trained models and achieves a Word Error Rate (WER) of 4% less than that reported as the benchmark on the same Urdu corpus using HMM. The decrease in WER after incorporating SSL is more significant with an increased validation data size.

In this study, the authors modify the mapping function by introducing a non‐linear term based on Riemann–Liouville definition of fractional derivatives and its application to the mean squared error in addition to the first‐order partial derivatives; thus create fractional variants of the least mean square (LMS) algorithm and its normalised version. The introduction of fractional term helps increase the convergence rate through the non‐linear update term which depends on the fractional order and the in‐process LMS weights; for steep gradient of the error measure, large changes are made to the weights which help the equaliser filter to better track the effects of multipath fading channels. They verify and validate the working of the proposed technique in the decision feedback equalisation of multipath fading channels for higher‐order quadrature amplitude modulations; comparative results are shown in terms of performance metrics as symbol‐error rate for various fractional orders and step sizes, combined equaliser and channel responses for different number of training symbols; simulations show that the proposed approach outperforms the conventional counterpart.