Georg Stemmer

Bio: Georg Stemmer is an academic researcher from Intel. The author has contributed to research in topics: Word error rate & Language model. The author has an hindex of 19, co-authored 76 publications receiving 6165 citations. Previous affiliations of Georg Stemmer include University of Erlangen-Nuremberg & Siemens.

The Kaldi Speech Recognition Toolkit

Abstract: We describe the design of Kaldi, a free, open-source toolkit for speech recognition research. Kaldi provides a speech recognition system based on finite-state automata (using the freely available OpenFst), together with detailed documentation and a comprehensive set of scripts for building complete recognition systems. Kaldi is written is C++, and the core library supports modeling of arbitrary phonetic-context sizes, acoustic modeling with subspace Gaussian mixture models (SGMM) as well as standard Gaussian mixture models, together with all commonly used linear and affine transforms. Kaldi is released under the Apache License v2.0, which is highly nonrestrictive, making it suitable for a wide community of users.

Language model modification for local speech recognition systems using remote sources

Abstract: A language model is modified for a local speech recognition system using remote speech recognition sources. In one example, a speech utterance is received. The speech utterance is sent to at least one remote speech recognition system. Text results corresponding to the utterance are received from the remote speech recognition system. A local text result is generated using local vocabulary. The received text results and the generated text result are compared to determine words that are out of the local vocabulary and the local vocabulary is updated using the out of vocabulary words.

Revising Perceptual Linear Prediction (PLP).

Abstract: Mel Frequency Cepstral Coefficients (MFCC) and Perceptual Linear Prediction (PLP) are the most popular acoustic features used in speech recognition. Often it depends on the task, which of the two methods leads to a better performance. In this work we develop acoustic features that combine the advantages of MFCC and PLP. Based on the observation that the techniques have many similarities, we revise the processing steps of PLP. In particular, the filter-bank, the equal-loudness pre-emphasis and the input for the linear prediction are improved. It is shown for a broadcast news transcription task and a corpus of children’s speech that the new variant of PLP performs better than both MFCC and conventional PLP for a wide range of clean and noisy acoustic conditions.

Detection of persons with Parkinson's disease by acoustic, vocal, and prosodic analysis

Abstract: 70% to 90% of patients with Parkinson's disease (PD) show an affected voice. Various studies revealed, that voice and prosody is one of the earliest indicators of PD. The issue of this study is to automatically detect whether the speech/voice of a person is affected by PD. We employ acoustic features, prosodic features and features derived from a two-mass model of the vocal folds on different kinds of speech tests: sustained phonations, syllable repetitions, read texts and monologues. Classification is performed in either case by SVMs. A correlation-based feature selection was performed, in order to identify the most important features for each of these systems. We report recognition results of 91% when trying to differentiate between normal speaking persons and speakers with PD in early stages with prosodic modeling. With acoustic modeling we achieved a recognition rate of 88% and with vocal modeling we achieved 79%. After feature selection these results could greatly be improved. But we expect those results to be too optimistic. We show that read texts and monologues are the most meaningful texts when it comes to the automatic detection of PD based on articulation, voice, and prosodic evaluations. The most important prosodic features were based on energy, pauses and F0. The masses and the compliances of spring were found to be the most important parameters of the two-mass vocal fold model.

Adaptive training using simple target models [speech recognition applications]

Abstract: Adaptive training aims at reducing the influence of speaker, channel and environment variability on the acoustic models. We describe an acoustic normalization approach to adaptive training. Phonetically irrelevant acoustic variability is reduced at the beginning of the training procedure w.r.t. a set of target models. The set of target models can be a set of HMMs or a Gaussian mixture model (GMM). CMLLR is applied to normalize the acoustic features. The normalized data contains less unwanted variability and is used to generate and train the recognition models. Employing a GMM as a target model leads to a text-independent procedure that can be embedded into the acoustic front-end. On a broadcast news transcription task we obtain relative reductions in WER of 7.8% in the first recognition pass over a conventionally trained system and of 3.4% in the second recognition pass over a SAT-trained system.

Dropout: a simple way to prevent neural networks from overfitting

Abstract: Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different "thinned" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets.

Representation Learning with Contrastive Predictive Coding

Abstract: While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments.

Librispeech: An ASR corpus based on public domain audio books

Abstract: This paper introduces a new corpus of read English speech, suitable for training and evaluating speech recognition systems. The LibriSpeech corpus is derived from audiobooks that are part of the LibriVox project, and contains 1000 hours of speech sampled at 16 kHz. We have made the corpus freely available for download, along with separately prepared language-model training data and pre-built language models. We show that acoustic models trained on LibriSpeech give lower error rate on the Wall Street Journal (WSJ) test sets than models trained on WSJ itself. We are also releasing Kaldi scripts that make it easy to build these systems.

SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition

Abstract: We present SpecAugment, a simple data augmentation method for speech recognition. SpecAugment is applied directly to the feature inputs of a neural network (i.e., filter bank coefficients). The augmentation policy consists of warping the features, masking blocks of frequency channels, and masking blocks of time steps. We apply SpecAugment on Listen, Attend and Spell networks for end-to-end speech recognition tasks. We achieve state-of-the-art performance on the LibriSpeech 960h and Swichboard 300h tasks, outperforming all prior work. On LibriSpeech, we achieve 6.8% WER on test-other without the use of a language model, and 5.8% WER with shallow fusion with a language model. This compares to the previous state-of-the-art hybrid system of 7.5% WER. For Switchboard, we achieve 7.2%/14.6% on the Switchboard/CallHome portion of the Hub5'00 test set without the use of a language model, and 6.8%/14.1% with shallow fusion, which compares to the previous state-of-the-art hybrid system at 8.3%/17.3% WER.

X-Vectors: Robust DNN Embeddings for Speaker Recognition

Abstract: In this paper, we use data augmentation to improve performance of deep neural network (DNN) embeddings for speaker recognition. The DNN, which is trained to discriminate between speakers, maps variable-length utterances to fixed-dimensional embeddings that we call x-vectors. Prior studies have found that embeddings leverage large-scale training datasets better than i-vectors. However, it can be challenging to collect substantial quantities of labeled data for training. We use data augmentation, consisting of added noise and reverberation, as an inexpensive method to multiply the amount of training data and improve robustness. The x-vectors are compared with i-vector baselines on Speakers in the Wild and NIST SRE 2016 Cantonese. We find that while augmentation is beneficial in the PLDA classifier, it is not helpful in the i-vector extractor. However, the x-vector DNN effectively exploits data augmentation, due to its supervised training. As a result, the x-vectors achieve superior performance on the evaluation datasets.