Faz dizili mikrofonlarla belirlenen koordinatlardan ses seçimi
Loading...
Files
Date
2015
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Hatay Mustafa Kemal Üniversitesi
Access Rights
info:eu-repo/semantics/openAccess
Abstract
Ses ve/veya konuşma kaydetmede ortam koşulları oldukça belirleyici olan olumsuz etkilere sahiptir. Bu etkilerin bertaraf edilmesi birden çok mikrofon kullanılması sayesinde sağlanabilmektedir. Birden çok mikrofonun kullanımı, mikrofonların yerleşim biçimlerine göre ortam içerisindeki tüm sesleri duyabilme olanağı vermektedir. Ancak, ortamdaki tüm seslerin duyulabilmesi, tüm seslere hâkim olma anlamına gelmemektedir. Zira ortamda birden çok ses kaynağı bulunması ve/veya ortamın yankılama etkisinin yüksek olması gibi durumlarda sesler birbirine karışmakta, kimi sesler örtülmekte, geri planda kalmakta ve anlaşılamamaktadır. Bu durum genel olarak ''kokteyl partisi problemi'' olarak adlandırılmaktadır. Problemin çözümü ise dizisel işaret işleme tekniklerinin çoklu mikrofon kayıtlarına uygulanmasıdır. Bu çalışmada birden fazla kişinin aynı anda konuştuğu ortamdaki seslerden sadece belirlenen Konumdaki seslerin alınarak istenmeyen seslerin bastırılması hedeflenmiştir. Geometrik olarak farklı noktalara yerleştirilmiş 8 mikrofondan oluşturulan mikrofon dizisi yardımı ile analitik uzayda konumları bilinen 2, 3, 4, 5, 6 ve 8 kişilik konuşmacılardan ses kayıtları alınarak 6 farklı çalışma yapılmıştır. Farklı mikrofonlardan alınan ses sinyallerine uygun gecikme oranları hesaplanmış, filtrele, geciktir ve topla yöntemleri kullanılarak daha kaliteli bir ses sinyali inşa edilmiştir. Anahtar Kelimler: Mikrofon dizileri, geciktir ve topla, demetleme, ses kayıt, üst üste binen konuşmalar
Conditions of the setting can have quite decisive negative impacts on voice or conversation recording. These unwanted decisive effects can be avoided by using multiple microphones. The use of multiple microphones gives the chance of hearing every sound according to the placements of the microphones. However, hearing all the sound does not mean that having full control over all of them. Since when there are multiple sources of sound and/or there is a high resonance effect in the setting, each sound can mix one other, some of them can be covered up, remain at the background, and so they may not be understood. This situation is generally called cocktail party problem. The solution for this is the application of sequential signal processing techniques on the recordings of multiple microphones. In this study, it is aimed to cover the sound that are differentiated from the other sound which come from only pre-determined directions in the setting when more than one person talk simultaneously. With the help of microphone array which consists of 8 microphones situated at different locations geometrically, 6 different studies have been conducted after taking voice recordings of groups of 2, 3, 4, 5, 6 and 8 people respectively, of whose locations are known on analytic spaces. Appropriate to audio signals taken from different microphones, lapse rates have been calculated; by using filter, delay and sum, beamforming techniques, more qualified audio signals have been created.
Conditions of the setting can have quite decisive negative impacts on voice or conversation recording. These unwanted decisive effects can be avoided by using multiple microphones. The use of multiple microphones gives the chance of hearing every sound according to the placements of the microphones. However, hearing all the sound does not mean that having full control over all of them. Since when there are multiple sources of sound and/or there is a high resonance effect in the setting, each sound can mix one other, some of them can be covered up, remain at the background, and so they may not be understood. This situation is generally called cocktail party problem. The solution for this is the application of sequential signal processing techniques on the recordings of multiple microphones. In this study, it is aimed to cover the sound that are differentiated from the other sound which come from only pre-determined directions in the setting when more than one person talk simultaneously. With the help of microphone array which consists of 8 microphones situated at different locations geometrically, 6 different studies have been conducted after taking voice recordings of groups of 2, 3, 4, 5, 6 and 8 people respectively, of whose locations are known on analytic spaces. Appropriate to audio signals taken from different microphones, lapse rates have been calculated; by using filter, delay and sum, beamforming techniques, more qualified audio signals have been created.
Description
Keywords
Elektrik ve Elektronik Mühendisliği, Electrical and Electronics Engineering, Microphone arrays, delay and sum, beamforming, voice recording, Overlapping Speech
