Abstract

This report introduces a speech database for multi-pitch tracking and describes the main steps of its development. The database contains the audio recordings and laryngograph signals of 20 English native speakers as well as the extracted pitch signals as a reference. The text material consists of 2342 “phonetically rich” sentences which are taken from the existing TIMIT corpus. Each of them was at least once read by a female and a male speaker. In total this database consists of 4720 recorded sentences. All recordings were supervised and carried out at the recording studio of the Institute of Broadband Communication at Graz University of Technology. The speech recording program and the appropriate recording setup will be described in detail in this report. Furthermore a snapshot of the data is provided by depicting some examples to show the connections between speech signals and pitch tracks. As this project will be continued, this report gives also a preview of the next steps like documentation, validation and distribution.

Full Text and additional Material

You can download the full version of the diploma thesis here.