Sound Event Detection Using Duration Robust Loss Function
Authors:
Daichi Akiyama,
Keisuke Imoto,
Noriyuki Tonami,
Yuki Okamoto,
Ryosuke Yamanishi,
Takahiro Fukumori,
Yoichi Yamashita
Abstract:
Many methods of sound event detection (SED) based on machine learning regard a segmented time frame as one data sample to model training. However, the sound durations of sound events vary greatly depending on the sound event class, e.g., the sound event ``fan'' has a long time duration, while the sound event ``mouse clicking'' is instantaneous. The difference in the time duration between sound eve…
▽ More
Many methods of sound event detection (SED) based on machine learning regard a segmented time frame as one data sample to model training. However, the sound durations of sound events vary greatly depending on the sound event class, e.g., the sound event ``fan'' has a long time duration, while the sound event ``mouse clicking'' is instantaneous. The difference in the time duration between sound event classes thus causes a serious data imbalance problem in SED. In this paper, we propose a method for SED using a duration robust loss function, which can focus model training on sound events of short duration. In the proposed method, we focus on a relationship between the duration of the sound event and the ease/difficulty of model training. In particular, many sound events of long duration (e.g., sound event ``fan'') are stationary sounds, which have less variation in their acoustic features and their model training is easy. Meanwhile, some sound events of short duration (e.g., sound event ``object impact'') have more than one audio pattern, such as attack, decay, and release parts. We thus apply a class-wise reweighting to the binary-cross entropy loss function depending on the ease/difficulty of model training. Evaluation experiments conducted using TUT Sound Events 2016/2017 and TUT Acoustic Scenes 2016 datasets show that the proposed method respectively improves the detection performance of sound events by 3.15 and 4.37 percentage points in macro- and micro-Fscores compared with a conventional method using the binary-cross entropy loss function.
△ Less
Submitted 26 June, 2020;
originally announced June 2020.
A Multi-Line Ammonia Survey of the Galactic Center Region with the Tsukuba 32-m Telescope - I. Observations and Data
Authors:
Hitoshi Arai,
Makoto Nagai,
Shinji Fujita,
Naomasa Nakai,
Masumichi Seta,
Aya Yamauchi,
Hiroyuki Kaneko,
Kenzaburo Hagiwara,
Koh-ichi Mamyoda,
Yusuke Miyamoto,
Masa-aki Horie,
Shun Ishii,
Yusuke Koide,
Mitsutoshi Ogino,
Masaki Maruyama,
Katsuaki Hirai,
Wataru Oshiro,
Satoshi Nagai,
Daiki Akiyama,
Keita Konakawa,
Hiroaki Nonogawa,
Dragan Salak,
Yuki Terabe,
Yoshiki Nihonmatsu,
Fumiyoshi Funahashi
Abstract:
We present survey data of the NH3 (J, K) = (1, 1)--(6, 6) lines, simultaneously observed with the Tsukuba 32-m telescope, in the main part of the central molecular zone of the Galaxy. The total number of on-source positions was 2655. The lowest three transitions were detected with S/N > 3 at 2323 positions (93% of all the on-source positions). Among 2323, the S/N of (J, K ) = (4, 4), (5, 5), and (…
▽ More
We present survey data of the NH3 (J, K) = (1, 1)--(6, 6) lines, simultaneously observed with the Tsukuba 32-m telescope, in the main part of the central molecular zone of the Galaxy. The total number of on-source positions was 2655. The lowest three transitions were detected with S/N > 3 at 2323 positions (93% of all the on-source positions). Among 2323, the S/N of (J, K ) = (4, 4), (5, 5), and (6, 6) exceeded 3.0 at 1426 (54%), 1150 (43%), and 1359 (51%) positions, respectively. Simultaneous observations of the lines enabled us to accurately derive intensity ratios with less systematic errors. Boltzmann plots indicate there are two temperature components: cold ($\sim$ 20 K) and warm ($\sim$ 100 K). Typical intensity ratios of Tmb(2,2)/Tmb(1,1), Tmb(4,4)/Tmb(2,2), Tmb(5,5)/Tmb(4,4), and Tmb(6,6)/Tmb(3,3) were 0.71, 0.45, 0.65, and 0.17, respectively. These line ratios correspond to diversity of rotational temperature, which results from mixing of the two temperature components.
△ Less
Submitted 30 August, 2016;
originally announced August 2016.
