CatBoost Machine Learning Based Feature Selection for Age and Gender Recognition in Short Speech Utterances

Badr, Ameer; Abdul-Hassan, Alia K.

doi:10.22266/ijies2021.0630.14

Cited by 10 publications

(9 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…After training to generate a basic learner, this gradient estimate variation results in the issue of overfitting. conventional techniques' gradient estimation bias is minimized by the CatBoost algorithm, which utilizes ordered boosting in place of the conventional method [24].…”

Section: Catboost Overviewmentioning

confidence: 99%

Game Theory-Infused Hybrid CatBoost- Extreme Learning Machine model for Reliable Identification of Rice Leaf Diseases for Advancing Agricultural Surveillance

Pratap,

Kumar

2024

Preprint

View full text Add to dashboard Cite

The global economy greatly relies on rice cultivation, yet the agricultural sector is primarily challenged by the prevalence of rice leaf diseases. This research introduces a novel Game Theory-Infused Hybrid CatBoost-Extreme Learning Machine (GT-CBELM) model tailored for the accurate and dependable detection of rice leaf diseases, thereby advancing agricultural surveillance practices. The proposed methodology harnesses cutting-edge image pre-processing methods, Contrast-Limited Adaptive Histogram Equalization (CLAHE), to enhance image quality and reveal critical disease-related details. The Grab Cut algorithm, achieves refined segmentation of disease-affected regions, leading to focused feature extraction and a substantial improvement in disease classification accuracy. Texture-specific features are extracted using the Grey Level Cooccurrence Matrix (GLCM) technique, effectively capturing essential structural information from affected areas. A groundbreaking contribution lies in the integration of game theory-based feature selection, empowered by the Banzhaf power index, which adeptly identifies relevant features while accounting for their inherent interdependencies, mitigating overfitting concerns and enhancing generalization capabilities. By seamlessly merging Game Theory with CatBoost algorithm’s robust categorical feature handling and ELM's pattern recognition process, the hybrid model excels in classifying three distinct rice leaf diseases brown spot, bacterial leaf blight, and leaf smut with remarkable precision and reliability. This innovative approach holds great promise for revolutionizing agricultural management strategies by enabling immediate and accurate disease identification, thereby contributing to enhanced crop health and agricultural productivity.

show abstract

Section: Catboost Overviewmentioning

confidence: 99%

Game Theory-Infused Hybrid CatBoost- Extreme Learning Machine model for Reliable Identification of Rice Leaf Diseases for Advancing Agricultural Surveillance

Pratap,

Kumar

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Et.al. [22,23] used the accumulated statistic of MFCC and LPC with their first and second derivatives, Spectral Sub-band Coefficients (SSC), as well as the first 4 formants (f1-f4), and MAE of 10.3 and 9.25 years on the VoxCeleb dataset, and 7.73 and 4.96 for male and female age on the TIMIT dataset.…”

Section: Related Workmentioning

confidence: 99%

“…Finding the ideal feature set to represent various bodily features, nevertheless, is challenging. In the research on speaker profiling, several feature selection and dimensionality reduction techniques have been used for age feature selection including Catbthe oost optimization technique [23], Principle Component Analysis (PCA) [3], and Linear Discriminant Analysis (LDA) [22,24,25].…”

Section: Related Workmentioning

confidence: 99%

Ensemble Feature Selection for Age Estimation from Speech

2023

IJCCCE

View full text Add to dashboard Cite

The voice signal carries a wide range of data about the speaker, including their physical characteristics, feelings, and level of health. There are several uses for the estimate of these physical characteristics from the speech in forensics, security, surveillance, marketing, and customer service. The primary goal of this research is to identify the auditory characteristics that aid in estimating a speaker’s age. To this end, an ensemble feature selection model is proposed that selects the best features from a baseline acoustic feature vector for age estimation from speech. Using a feature vector that covers various spectral, temporal, and prosodic aspects of speech, an ensemble-based automatic feature selection is performed by, first calculating the feature importance or ranks based on individual feature selection methods, then voting is applied to the resulting feature ranks to attain the top-ranked subset by all feature selection methods. The proposed method is evaluated on the TIMIT dataset and achieved a mean absolute error (MAE) of 5.58 years and 5.12 years for male and female age estimation. Index Terms— Age Estimation, Feature Selection, Ensemble Selection, TIMIT dataset.

show abstract

“…The timedomain autocorrelation method is the most common method of use [56]. The autocorrelation approach computes the dot-product of the speech signal with a shifted version of that signal [57]. Studies of F0 in speaker profiling show F0 is inversely proportional to the height of a speaker [58] [53].…”

Section: Fundamental Frequency / Pitchmentioning

confidence: 99%

Review of Automatic Speaker Profiling: Features, Methods, and Challenges

Jaid,

Abdul Hassan

2023

Iraqi Journal of Science

View full text Add to dashboard Cite

Automatic Speaker Profiling (ASP), is concerned with estimating the physical traits of a person from their voice. These traits include gender, age, ethnicity, and physical parameters. Reliable ASP has a wide range of applications such as mobile shopping, customer service, robotics, forensics, security, and surveillance systems. Research in ASP has gained interest in the last decade, however, it was focused on different tasks individually, such as age, height, or gender. In this work, a review of existing studies on different tasks of speaker profiling is performed. These tasks include age estimation and classification, gender detection, height, and weight estimation This study aims to provide insight into the work of ASP, available datasets, feature extraction techniques, and learning models. Further, the performance of current speaker profiling systems is investigated. Finally, the challenges of speaker profiling are presented at the end of this review.

show abstract

CatBoost Machine Learning Based Feature Selection for Age and Gender Recognition in Short Speech Utterances

Cited by 10 publications

References 0 publications

Game Theory-Infused Hybrid CatBoost- Extreme Learning Machine model for Reliable Identification of Rice Leaf Diseases for Advancing Agricultural Surveillance

Game Theory-Infused Hybrid CatBoost- Extreme Learning Machine model for Reliable Identification of Rice Leaf Diseases for Advancing Agricultural Surveillance

Ensemble Feature Selection for Age Estimation from Speech

Review of Automatic Speaker Profiling: Features, Methods, and Challenges

Contact Info

Product

Resources

About