Signal Processing - May 2016 - 7

rate reduction is shown for phone
re cog nition and large vocabulary
voice search.
October 2014

Robust Sound Event Classification
Using Deep Neural Networks
McLoughlin, I.; Zhang, H.; Xie, Z.;
Song, Y.; Xiao, W.
This paper outlines a sound event classification framework that compares
auditory image front end features with
spectrogram image-based front-end
features, using a support vector machine and deep neural network classifiers. Performance is evaluated on a
standard robust classification task in
different levels of corrupting noise,
and with several system enhancements.
March 2015

Keyword Extraction and
Clustering for Document
Recommendation in Conversations
Habibi, M.; Popescu-Belis, A.
The authors extract key words from an
automatic speech recognition system
and use a topic model that favors
diversity in the key word set and derive
topicaly separated queries to run on
Wikipedia. It attempts to maximize the
probability of making at least one relevant recognition. Testing with human
judges shows improvement over word
frequency or topic similarity methods.
April 2015

A Regression Approach to Speech
Enhancement Based on
Deep Neural Networks
Xu, Y.; Du, J.; Dai, L.-R.; Lee, C.-H.
This paper finds a map from a noisy
speech signal to a clean one using
deep neural networks. Dropout and
noise-aware training strategies lead to
robust performance to highly nonstationary noise.
January 2015

From Feedforward to Recurrent
LSTM Neural Networks for
Language Modeling
Sundermeyer; M.; Ney, H.;
Schluter, R.
This paper compares count models
to feedforward, recurrent, and long

image licensed by ingram publishing

short-term memory neural network
variants on two large-vocabulary
speech recognition tasks. The in creased computational complexity
requires efficient search methods for
the neural networks. Performance is
evaluated in terms of perplexity and
word error rate.
March 2015

use of acoustic models trained jointly
with the same feature enhancement or
model adaptation process used in the
testing stage.
April 2014

Single Frequency Filtering
Approach for Discriminating
Speech and Nonspeech

Aneeja, G.; Yegnanarayana, B.
The mean and variance of the noisecompensated weighted envelopes are
Sentence Compression for
computed across frequency at each time
Aspect-Based Sentiment Analysis
instant. Because the variance of the
Che, W.; Zhao, Y.; Guo, H.; Su, Z.; Liu, T.
spectral information across frequency is
The method described in this paper
higher for speech
compresses sentences by
and lower for many
removing infor mation
This issue's "reader's
types of noises, the
unrelated to sentiment
choice" column lists
method obtains betusing a conditional ranthe top ten papers
ter performance
dom field model. The
most downloaded for
than adaptive multishorter sentences are
rate VAD2.
easier to parse allowing
the past year.
April 2015
fine-grained aspect-based
sentiment analysis.
December 2015
Voice Conversion Using RNN

Pre-Trained by Recurrent Temporal
Restricted Boltzmann Machines

An Overview of Noise-Robust
Automatic Speech
Recognition
Li, J.; Deng, L.; Gong, Y.;
Haeb-Umbach, R.
This paper provides a thorough overview of modern noise-robust techniques
for ASR developed over the past 30
years. Noise robust techniques are
analyzed using five criteria: 1) featuredomain versus model-domain processing, 2) the use of prior knowledge
about the acoustic environment distortion, 3) the use of explicit environmentdistortion models, 4) deterministic
versus uncertainty processing, and 5) the
IEEE Signal Processing Magazine

May 2016

Nakashika, T.; Takiguchi, T.; Ariki, Y.
This paper converts voices using one
recurrent temporal restricted Bolt zmann machine for the source and
one for the destination speaker to
convert both speaker signals to the
model parameters, then a neural network to convert between the models'
parameters. This method compares
favorably with Gaussian mixture
model methods.
March 2015

sP
7

Table of Contents for the Digital Edition of Signal Processing - May 2016