Japanese Speech Database Research Articles

English Read by Japanese Phonetic Corpus: An Interim Report

The primary purpose of this paper is to explain the procedure of developing the English Read by Japanese Phonetic Corpus. A series of preliminary studies (Makino 2007, 2008, 2009) made it clear that a phonetically-transcribed computerized corpus of Japanese speakers’ English speech was worth making. Because corpus studies on L2 pronunciation have been very rare, we intend to fill this gap. For the corpus building, the 1,902 sentence files in the English Read by Japanese speech database scored for their individual sounds by American English teachers trained in phonetics in Minematsu, et al. (2002b) have been chosen. The files were pre-processed with the Penn Phonetics Lab Forced Aligner to generate Praat TextGrids where target English words and phonemes were forced-aligned to the speech files. Two additional tiers (actual phones and substitutions) were added to those TextGrids, the actual phones were manually transcribed and the other tiers were aligned to that tier. Then the TextGrids were imported to ELAN, which has a much better searching functionality. So far, fewer than 10% of the files have been completed and the corpus-building is still in its initial stage. The secondary purpose of this paper is to report on some findings from the small part of the corpus that has been completed. Although it is still premature to talk of any tendency in the corpus, it is worth noting that we have found evidence of phenomena which are not readily predicted from L1 phonological transfer, such as the spirantization of voiceless plosives, which is not considered normal in the pronunciation of Japanese.

Optimization of target cost weights in concatenative speech synthesis with very short segments of 5‐ms duration

If a concatenative speech synthesis system uses more short speech segments, it increases the potential to generate natural speech because the concatenation variation becomes greater. Recently, a synthesis approach was proposed in which very short (5 ms) segments are used [T. Hirai and S. Tenpaku, 5th ISCA Speech Synthesis Workshop, pp. 37–42 (2004), http://www.ssw5.org/]. In that approach, the target cost (how close a database segment is to desired segment) was defined as the simple average of the root mean squares (rms’s) of the difference between the features of the database segment and the desired segment for simplicity. Therefore, it has been expected to optimize the weight of each rms. A Japanese speech database was used to optimize the weights and to evaluate its effects. For corpus construction, 150 utterances were selected from the database. Ten other utterances were selected randomly for generation of feature time series as natural targets for synthesis. Half of them were used for optimization using many weight combinations in synthesis to determine the optimal weight set that shows the minimum concatenation distortion. Distortion for the other half of the utterance synthesis with optimized weights was reduced 34.4% compared to the former approach.

The present status, progress, and usage of speech databases in Japan

The present status, progress and usage of Japanese speech database has been described. The database project in Japan started in the early 1980s. The first was by the Japan Electronic Industry Development Association (JEIDA), which aimed at creating a speech database to evaluate performance of the existing speech input/output machines and systems. Several database projects have been undertaken since then, including the one initiated by the Advanced Telecommunication Research Institute (ATR), and now we have reached a point where an enormous amount of spontaneous speech data is available. A survey was conducted recently on usage of the presently existing speech databases among industry and university institutions in Japan where speech research is now actively going on. It was revealed that the ATR’s continuous speech database is the most frequently used, followed by the equivalent version from the Acoustical Society of Japan.

Creation and evaluation of a large-scale Japanese speech database

Creation and evaluation of a large-scale Japanese speech database

Phoneme recognition with elliptic discrimination neural units

AbstractMany researchers achieved high phoneme recognition rates by multilayered neural networks with linear discrimination neural (LDN) units. However, it is difficult to analyze which components of the input are important to each unit in those LDN networks.This paper proposed a multilayer neural network with elliptic discrimination neural (EDN) units so that the functions of each unit in the network may be interpreted more definitely. The center of the elliptic discrimination boundary of a neural unit corresponds to a typical point in an input space. The radii of the ellipse express the extent of the corresponding components in the input space, hence it becomes clear which components of the input space are important to each unit in the EDN network.To compare the performance of EDN and LDN networks, recognition experiments of phonemes /b, d, g/ in 5240 tokens of a Japanese speech database were carried out. In the experiments, recognition rates were obtained by EDN networks as high as the rate by an LDN network. Also, it was confirmed which components of the input space are important to each unit in the EDN network.

ATR Japanese speech database as a tool of speech recognition and synthesis

A large-scale Japanese speech database has been described. The database basically consists of (1) a word speech database, (2) a continuous speech database, (3) a database for a large number of speakers, and (4) a database for speech synthesis. Multiple transcriptions have been made in five different layers from simple phonemic descriptions to fine acoustic-phonetic transcriptions. The database has been used to develop algorithms in speech recognition and synthesis studies and to find acoustic, phonetic and linguistic evidence that will serve as basic data for speech technologies.

A Japanese speech database for various kinds of research purposes

A Japanese speech database for various kinds of research purposes