Descriptive Statistical Evaluation of Zero Crossing Rate (ZCR) Features in Indonesian and Buginese Speech Signals for Speech Processing Applications

Sriwijanaka Yudi Hartono; Saidah Suyuti; Muhammad Iqbal

doi:10.31963/elekterika.v23i1.6450

Authors

Sriwijanaka Yudi Hartono Universitas Muslim Indonesia
Saidah Suyuti Universitas Muslim Indonesia
Muhammad Iqbal Universitas Muslim Indonesia

DOI:

https://doi.org/10.31963/elekterika.v23i1.6450

Keywords:

algoritma, zero crossing rate, staistika, descriptive, buginese

Abstract

This study analyzes the statistical characteristics of the Zero Crossing Rate (ZCR) in Indonesian and Buginese speech signals produced by both male and female speakers. The objectives of this research are to identify the statistical patterns of ZCR, compare the distribution of its key descriptive measures between the two languages, and provide a scientific foundation for developing speech processing systems that are adaptive to regional languages. Three types of utterances were examined, and ZCR features were extracted using descriptive statistical parameters including mean, variance, standard deviation, skewness, and kurtosis. The findings show that Buginese speech exhibits higher and more variable ZCR values compared to Indonesian speech. A statistical summary from all tables indicates that the mean ZCR of Buginese speech ranges from 0.0727 to 0.2464, whereas Indonesian speech ranges from 0.0338 to 0.1577. The variance in Buginese speech reaches 0.009023–0.022207, significantly higher than Indonesian speech, which ranges from 0.000738 to 0.012896. The standard deviation of Buginese speech reaches 0.149019, while Indonesian speech ranges from 0.027173 to 0.112889. These wider distributions in Buginese speech reflect more intense polarity fluctuations, influenced by its distinctive phonetic features such as geminate consonants, sharp fricatives, and glottal stops. In contrast, Indonesian speech shows more stable ZCR characteristics, with narrower distributions of skewness and kurtosis, consistent with its dominance of vowels and softer voiced consonants. These statistical differences are consistent across both male and female speakers, although female speakers tend to exhibit slightly more stable patterns. The results confirm that ZCR is an effective acoustic feature for distinguishing the speech characteristics of Indonesian and Buginese. Overall, this study provides a strong scientific basis for the development of speech processing and automatic speech recognition (ASR) systems that are more adaptive to the phonetic characteristics of regional languages.

References

Yudi Hartono S, Suyuti S. Method of extracting speech characteristics of Bugis regional language. J Teknol Elektron. 2025;22(1):35–40.

[2] Yudi Hartono S, Suyuti S. Ekstraksi pitch sinyal wicara (speech) bahasa daerah Bugis dengan algoritma autokorelasi. J INSTEK. 2024;9(2):125–54.

[3] Hartono YS, Basalamah A. Ekstraksi pitch pada suara laki-laki dan perempuan dengan metode algoritma autokorelasi dan AMDF untuk penentuan tipe jenis jangkauan suara berdasarkan tipe suara. Penelitian Dosen Intern PUF, Fakultas Teknik, Universitas Muslim Indonesia; 2016.

[4] Sun Y, Zhang Z, Schuller BW. On the analysis of speech emotion recognition using deep spectrum features with LSTM networks. IEEE Access. 2020;8:165811–23.

[5] Mohd Hanifa, R., Isa, K., Mohamad, S., Shah, S. M., Nathan, S. S., Ramle, R., & Berahim, M. (2024). Voiced and unvoiced separation in Malay speech using zero crossing rate and energy. Indonesian Journal of Electrical Engineering and Computer Science:16(2):775-780.

[6] Endah, S. N. et al. (2022). Continuous speech segmentation menggunakan dynamic thresholding pada fitur short-term (ZCR, energy, spectral flux); akurasi segmentasi kata mencapai 97,5 %, Jurnal of Engineering Science and Technology. 2022.:4(7):2719-2935

[7] Gunawan A, Riza LS, Arifianto A. Speech recognition system using MFCC and Dynamic Time Warping for Bahasa Indonesia. J Ilmu Komput dan Inform. 2021;14(1):37–44.

[8] Chen J, Wang Q, Xiao X, et al. End-to-end speech emotion recognition using 1D convolutional long short-term memory networks. Int J Speech Technol. 2021;24:695–704.

[9] Pardede HF, Adhi P, Zilvan V, Ramdan A, Krisnandi D. Deep convolutional neural networks-based features for Indonesian large vocabulary speech recognition. International Journal of Artificial Intelligence (IJ-AI). 2023;12(2):610–oi 10.11591/ijai.v12.i2.pp610-61710.

[10] Dura M, Kadiri MS, Diouri O. An improved voice command recognition system using MFCC and deep learning. Procedia Comput Sci. 2020;170:718–23.

[11] Putra WR, Amalia F. Analisis ciri akustik vokal Bahasa Bugis dan Bahasa Indonesia menggunakan metode Zero Crossing Rate dan Spectral Centroid. J Ilm Teknol Inform. 2023;9(2):150–9.

[12] Rahman A, Nugroho A. Deteksi gender suara manusia menggunakan metode Zero Crossing Rate dan algoritma k-NN. J Fotonik. 2022;10(1):11–8.

[13] Nugraha R, Setyawan I. Penerapan ekstraksi ciri ZCR dan Energy untuk klasifikasi suara manusia. J Teknol dan Sistem Komput. 2022;10(3):321–7.

[14] Huang Z, Epps J, He L. An investigation of cross-language speech emotion recognition. Speech Commun. 2020;120:20–36.

[15] Putri DM, Siregar A. Klasifikasi suara berdasarkan jenis kelamin menggunakan algoritma Zero Crossing Rate dan Naive Bayes. J Teknol Inf dan Komun. 2023;7(1):75–81.

[16] Arsyad M, Nurfadillah S. Implementasi Zero Crossing Rate dan Energy untuk identifikasi ujaran frustasi. J Rekayasa Elektron dan Komput. 2024;5(2):88–93.

[17] Iskandar Y, Pratama S. Analisis karakteristik sinyal wicara berbasis bahasa daerah menggunakan metode MFCC dan ZCR. J Teknol Informasi. 2023;12(2):141–8.

[18] Latief R, Salam U. Pengenalan ucapan bahasa Bugis dengan metode deep learning berbasis fitur ZCR dan MFCC. J Teknol Informasi dan Ilmu Komputer. 2025;8(1):101–8.

[19] Safitri R, Wirawan B. Klasifikasi emosi suara Bahasa Indonesia menggunakan CNN dan fitur ZCR. Jurnal Sains Komputer. 2023;4(1):23–31.

[20] Lee J., Kim S.,Statistical Analysis of Acoustic Features Including Zero Crossing Rate for Speaker Identification IEEE Acess,2021;9;14567-14577.