Takahashi Toru

This paper describes development and evaluation of a notification system of tram position by using global positioning system. We also show a design concept of the system. The most important point of the concept is that the system is constructed from easily acquirable and general purpose equipments since we intend to promote a use of location notification system in other transportations, such as bus, taxi, and train. A prototype system based on the concept is developed. It is evaluated on Hankai tram in service. We test two map matching algorithm to reduce estimation error and optimize length between anchor points. We found that a suitable period of measuring location is 1 or 2 seconds. Experimental results show that an expected total delay for showing location is 3 second and maximum error of the location is 100m. It is confirmed that we can construct a location notification system from easily acquirable and general purpose equipments.

本稿では，歌声と朗読音声を識別するシステムについて述べる．入力は無雑音音声，出力は歌声と朗読音声それぞれの尤度 （連続値） である．従来，スペクトル包絡 （MFCC） と基本周波数 （F0） の時間変化に基づいた識別システムが報告されている．これらの特徴量に基づく識別器に，スペクトル変化量のピーク間隔という，音素継続時間に関連する特徴量に基づく識別器を加え，入力音声長に応じて各識別器への重みを変化させた．実験の結果，従来システムでは1秒の音声に対し 86.7% の精度であったのに対し，本システムでは 90.2% という結果を得た．本システムが実時間で動作するデモアプリケーションについても述べる．In this paper we describe a system that discriminates between singing and speaking voices. Given a clean speech signal, it outputs the likelihood of each of the singing and speaking voices. Previous systems use temporal transition of spectral envelope (MFCC) and fundamental frequency (F0) as discrimina- tion features. Our system adds peak interval of spectral change as a phoneme duration feature and weights these features according to the duration of the input speech signal. Experimental results with one-second speech signal show that our system achieves 90.2 % accuracy compared to 86.7 % with previous systems. We also describe a real-time application demonstrating our system.