Estimasi F0 Berbasis IFAS

Ahmada, Dipta Nusbangga Haikal (2018) Estimasi F0 Berbasis IFAS. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
2311040000083_Undergraduate-Theses.pdf - Published Version
Restricted to Repository staff only

Download (5MB) | Request a copy

Abstract

Pada penelitian tugas akhir ini, dilakukan eksperimental im- plementasi metode estimasi F 0 berbasis instantaneous frequency amplitude spectrum (IFAS) pada sinyal suara (speech). F 0 meru- pakan frekuensi yang berfungsi mengkarakterisasi struktur har- monik pada sinyal suara. Dalam implementasinya, F 0 merupakan frekuensi yang memaksimalkan fungsi evaluasi IFAS. Metode esti- masi F 0 berbasis IFAS merupakan metode estimasi F 0 yang cukup handal dan akurat jika dibandingkan metode lainnya, seperti: cepstrum dan autocorrelation. Di samping itu, dalam implemen- tasi estimasi F 0 berbasis IFAS, didefinisikan suatu besaran yang mendeskripsikan keberadaan struktur harmonik dalam sinyal suara secara kuantitatif, yang disebut sebagai harmonicity mea- sure. Dalam implementasinya, Harmonicity measure digunakan selektor yang memilih optimal interval frekuensi yang padanya fungsi evaluasi diterapkan, sehingga, error estimasi F 0 dapat direduksi secara substansial. Pada implementasi estimasi F 0 berbasis IFAS, meskipun har- monicity measure memberikan hasil yang optimal, namun keber- adaan representasi struktur harmonik masih belum jelas. Hal ini ditunjukkan pada hasil representasi struktur harmonik instanta- neous frequency (IF), dimana, pada transisi frekuensi harmonik satu ke frekuensi harmonik berikutnya menunjukkan adanya diskontinuitas. Selain memberikan struktur harmonik yang masih belum jelas, diskontinuitas pada representasi struktur harmonik, membatasi interval frekuensi evaluasi. Oleh karena itu, pada penelitian tugas akhir ini, diusulkan formulasi IF bebasis pada Gaussian window dengan formulasi IF yang dapat mereduksi diskontinuitas pada representasi struktur harmonik IF. Hasil dari IF Gaussian Window diperoleh representasi struktur harmonik yang lebih jelas. Pengujian dari hasil estimasi F 0 bebasis IFAS Gaussian window pada database suara yang terdiri dari 2 pria dan 2 wanita dengan 20 pengucapan, menunjukkan hasil fine pitch error pria sebesar 6.2934% dan wanita sebesar 4.9833% dibandingkan dengan metode STRAIGHT. =========== In this final project, experimental implementation of F 0 method based on instantaneous frequency estimation amplitude spectrum (IFAS) on the speech signal. F 0 is a frequency functioning to characterize the harmonic structure of the sound signal. In its implementation, F 0 is frequency that can maximizes IFAS evalua-tion function. The IFAS F 0 estimation method is a sufficient F 0 estimation method reliable and accurate when compared to other methods, such as: cepstrum and utocorrelation. In addition, in the implementation of an IFAS-based F 0 estimate, a value is defined to describe the existence of harmonic structure in the speech signal quantitatively, which is called the harmonicity measure. In its implementation, Harmonicity measure is used as selector that selects the optimal frequency interval on it evaluation function is applied, so, the estimation error F 0 can be reduced substantially. In the implementation of an IFAS-based F 0 estimate, although the harmonicity measure gives optimal results, the existence of a harmonic structure representation remains unclear. This matter shown in the result of the representation of the instantaneous fre- quency (IF) harmonic structure, where, at the harmonic frequency transition one to the next harmonic frequency indicates the presence discontinuity. In addition to give the harmonic structure is still unclear, discontinuities on harmonic structure representation, limiting the evaluation frequency interval. Therefore, on research of this final task, proposed IF formulation bebasis on Gaussian window with IF formulation that can reduce discontinuity on the representation of IF harmonic structure. Result of IF Gaussian Window obtained the representation of harmonic structure which is clearer. Testing of IFD F 0 bebasis estimation results Gaussian window on a sound database consisting of 2 men and 2 women with 20 pronunciations, showing fine pitch error for male is 6.2934% and for women is 4.9833% compared with the STRAIGHT method.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: F0; Gaussian Window; IF; IFAS
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5102.9 Signal processing.
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms
Divisions: Faculty of Industrial Technology > Physics Engineering
Depositing User: Dipta nusbangga haikal ahmada
Date Deposited: 05 Feb 2018 05:20
Last Modified: 05 Feb 2018 05:20
URI: http://repository.its.ac.id/id/eprint/51084

Actions (login required)

View Item View Item