替一个日本同事发帖，具体他的研究方向如下，感兴趣的可以把CV发我(l.tang@auckland.ac.nz) 或者直接联系他本人(y.hioka@auckland.ac.nz)：

Descriptions of current research projects
Dr. Yusuke Hioka, Senior Lecturer
Acoustics Research Centre, University of Auckland

1. Investigating effect of noise and room acoustics on the speech intelligibility of non-native listeners

Intelligibility of speech is challenged by adverse conditions including the clarity of the speakers’ pronunciation, environmental noise, and whether or not the listener is a non-native listener. Combination of these conditions make these problems worse; it severely affects the quality of verbal communication. This is a significant issue in today’s world where over half of the world population is multi-lingual. Despite that fact, the comprehension ability of non-native listeners immersed in a noisy environment is not well understood. To improve communication, we will reveal and quantify how the ambient acoustic environment of a listener affects the understanding of non-native listeners at different level of proficiency compared to that of native listeners by conducting subjective listening tests. Ultimate goal of this project is to establish solid guidelines of room acoustics as well as audio signal processing for helping non-native listeners have better communication experiences.

2. Speech enhancement for vulnerable listeners

Speech enhancement is an area that has been actively studied for many decades. Although various speech enhancement algorithms have been invented, only a little attention has been paid to the ability of the listeners of the enhanced speech; namely most state of the art techniques target at listeners with normal hearing who are the native speaker of the language spoken. This project will focus on developing novel speech enhancement algorithms for vulnerable listeners who are disadvantaged by their hearing ability and/or the proficiency of the language spoken.

3. Localisation of bird calls in NZ bush

Have you ever spotted kiwi birds in NZ bush? Can you imagine how many of them are living in a forest? It is actually quite rare that you come across birds like kiwis because of their declining population and nocturnal lifestyle. Therefore it is a quite challenging task for the rangers and researchers to study the ecology of such birds without actually staying in a forest for weeks and months. The DOC (Department of Conservation) has
been attempting to identify bird calls from audio recordings collected in a forest however
so far this has been done completely manually (hiring people and get them to literally
"listen and identify" bird calls) which is not cost effective and the results can also be
inaccurate. Recently the DOC called a group of researchers around NZ (known as
AviaNZ; https://www.avianz.net/) to develop an automated system to identify bird calls as
well as measuring abundance of species. Yusuke Hioka is a part of the group in charge
of extracting various acoustic information from the recording. In this project we focus on
localising bird calls (i.e. extracting "location" of birds from recordings) using digital signal
processing with microphone arrays (i.e. array of more than one microphones).

4. High quality sound recording and source localisation using unmanned aerial vehicles

Unmanned aerial vehicles (UAVs) have recently gained huge popularity across a wide
range of applications, including filming, search and rescue, and surveillance. Such
applications take advantage of capturing visual information (i.e. video and imagery) that
are otherwise impossible without making use of UAVs. On the other hand, audio signals
are also one that should not be overlooked. It is common to encounter environments
that are often remote and harsh, which can easily render visual information unusable.
This is not the case with audio. However, audio recording using UAVs have shown to be
challenging due to the high noise levels radiated from the UAV rotors. This significantly
affects the quality of the audio signals to aid with any application.
The problem has been tackled by the Acoustics Research Centre (ARC), UoA, for which
a UAV system with quiet rotors, equipped with an array of microphones and a signal
processing algorithm, was developed to effectively record desired audio in-flight while
removing the UAV rotor noise. Recently, a method based on machine learning was used
to explore possibilities of predicting UAV rotor noise with a hybrid of microphone and
non-acoustical information. However, a common problem with such data-driven system
is the lack of transparency between the inputs and the result it produces. To this end,
studies have been made to unravel these ambiguities with the help of analytical
modelling. This project will focus on incorporating these analytical findings to optimise
the current signal processing algorithm.

5. Design of real-time speech masking system

Speech masking is a technique being used to hide confidential information in a target
speech where a jammer sound (i.e. masker) is played to hinder understanding by the
human auditory system. Since the maskers could cause annoyance for listeners, the proposed research will identify a novel design of maskers that will NOT cause any psychological disruptions to the listeners while maintaining its masking performance. A signal processing algorithm for creating a masker, which is grounded on the expertise in acoustics, linguistics, and audiology will be developed and tested through subjective listening tests.

Recent relevant publications
[1] H. Masuda, Y. Hioka, J. James, and C. Watson, “Protecting speech privacy from native/non-native listeners - effect of masker type”, In International Congress on Phonetic Sciences (ICPhS), 3070– 3074, Aug 2019.
[2] Y. Koizumi, K. Niwa, Y. Hioka, K. Kobayashi, and Y. Haneda, “DNN-based source enhancement to increase objective sound quality assessment score”, IEEE/ACM Transactions on Audio, Speech and Language Processing, 26(10):1780–1792, October 2018.
[3-1] B. Ollivier, A. Pepperell, Z. Halstead, and Y. Hioka, “Noise robust bird call localisation using the generalised cross-correlation in the wavelet domain” Journal of the Acoustical Society of America (Accepted)
[3-2] A. Pepperell, Z. Halstead, B. Ollivier, and Y. Hioka, “Performance of sound source localisation for bird calls in native New Zealand bush”, New Zealand Acoustics, 32(2):15-24, 2019. [4-1] Y. Hioka, M. Kingan, G. Schmid, R. McKay, and K. Stol, “Design of an unmanned aerial vehicle mounted system for quiet audio recording”, Applied Acoustics, 155:423–427, December 2019. [4-2] B. Yen, Y. Hioka, and B Mace, “Improving power spectral density estimation of unmanned aerial vehicle rotor noise by learning from non-acoustic information”, In 16th International Workshop on Acoustic Signal Enhancement (IWAENC), 545–549, Sep 2018.
[4-3] B. Yen, Y. Hioka, and B Mace, “Estimating power spectral density of unmanned aerial vehicle rotor noise using multisensory information”, In 26th European Signal Processing Conference (EUSIPCO 2018), 2434–2438, Sep 2018.
[5-1] Y. Hioka, J. Tang, and J. Wan, “Effect of adding artificial reverberation to speech-like masking sound”, Applied Acoustics, 114:171–178, Dec 2016.
[5-2] Y. Hioka, J. James, and C.I. Watson, “Masker design for real-time informational masking with mitigated annoyance”, Applied Acoustics, (Accepted) 返回小木虫查看更多