Modified TIMIT, SITW, and NIST 2008 Datasets

The modified TIMIT, SITW, NIST 2008 Which used in the work (Adaptive Weight Control for DNN-GMM-Fuzzy Fusion in Robust Speaker Identification) In the modified databases, 120 speakers were selected from each of the TIMIT, SITW, NIST 2008, and NTIMIT databases to ensure consistency with the experimental protocols reported in [1] and [2]. Each speaker contributed 10 speech utterances, with two evaluation protocols considered: 6/4 (six utterances for training and four for testing) and 8/2 (eight utterances for training and two for testing). To maintain uniformity across all experiments, a fixed speech duration of 129,250 samples (approximately 8 s) was used for each utterance. When an utterance was shorter than the required duration, additional speech segments from the same speaker were concatenated to achieve the target length. This resulted in a total of 1,200 speech utterances (120 speakers × 10 utterances) for each modified database. [1] M. T. S. Al-Kaltakchi, W. L. Woo, S. Dlay, & J. A. Chambers, "Evaluation of a speaker identification system with and without fusion using three databases in the presence of noise and handset effects," EURASIP Journal on Advances in Signal Processing, vol. 2017, no. 1, p. 80, 2017. doi:10.1186/s13634-017-0515-7 [2] M. T. S. Al-Kaltakchi, M. A. M. Abdullah, W. L. Woo, & S. S. Dlay, "Combined i-vector and extreme learning machine approach for robust speaker identification and evaluation with SITW 2016, NIST 2008, TIMIT databases," Circuits, Systems, and Signal Processing, vol. 40, no. 10, pp. 4903–4923, 2021. doi:10.1007/s00034-021-01697-7