Spoken language identification using i-vectors, x-vectors, PLDA and logistic regression

Abstract

In this paper, i-vector and x-vector is used to extract the features from speech signal from local Indonesia languages, namely Javanese, Sundanese and Minang languages to help classifier identify the language spoken by the speaker. Probabilistic linear discriminant analysis (PLDA) are used as the baseline classifier and logistic regression technique are used because of prior studies showing logistic regression has better performance than PLDA for classifying speech data. Once these features are extracted. The feature is going to be classified using the classifier mentioned before. In the experiment, we tried to segment the test data to three segment such as 3, 10, and 30 seconds. This study is expanded by testing multiple parameters on the i-vector and x-vector method then comparing PLDA and logistic regression performance as its classifier. The x-vector has better score than i-vector for every segmented data while using PLDA as its classifier, except where the i-vector and x-vector is using logistic regression, i-vector still has better accuracy compared to x-vector.

Performance analysis of demand forecasting in energy consumption based on ensemble model

“…From the above result, the decision boundary has been calculated for threshold of returning a probability score between 0 and 1 [36]. Finally, the cost function represents the optimization of convex function to minimize the cost value and finding the global minimum of logistic regression.…”

Artificial speech detection using image-based features and random forest classifier

“…In speaker verification, the universal background model (UBM) is a speaker model that represents broad attributes and characteristics that can be compared to the specific person being verified [1]. Later, i-vector and x-vector [2] based ASV systems were introduced to replace the gaussian mixture model-universal background model (GMM-UBM) based ASV systems. Deep learning approaches [3] such as recurrent neural network (RNN) [4] as a backend classifier was shown the capability in speaker verification with a low equal error rate (EER).…”

Spoken language identification on 4 Indonesian local languages using deep learning

“…Many studies on language identification have been conducted, with various feature extraction and classification techniques being used. Several techniques are used to extract features from the audio data, including phone recognition followed by language modeling (PRLM) [5] and parallel phone recognition followed by language modeling (PPRLM) [5] for phonetic approach or perceptual linear prediction (PLP) [5], mel-frequency cepstral coefficient (MFCC) [6]- [8], i-vector [8], [9] and x-vector [10] for the acoustic approx neural networks [11], convolutional neural networks (CNN) [12], [13], logistic regression (LR) [8], PLDA [14], Gaussian mixture model (GMM) [15], [16], support vector machine [17], [18] are among techniques used to classify the language spoken.…”

“…These findings were obtained using a dataset of speech corpora in three Indonesian local languages (Javanese, Sundanese, and Minangkabau) that were independently recorded. Abdurrahman et al [8] used an acoustic approach with ivector and x-vector extraction features with probabilistic linear discriminant analysis (PLDA) and LR classifications to study three Indonesian local languages. As a result, the x-vector performs best when using PLDA, while the i-vector outperforms the x-vector when using LR.…”

“…The author wishes to expand on the work of Safitri et al [5], Abdurrahman et al [8], and Wicaksana et al [7]. The proposed study will employ four Indonesian dialects: Javanese, Sundanese, Minangkabau, and Buginese.…”