What question did this study set out to answer?

The aim is to investigate a hybrid deep learning model's effectiveness for classifying gender from Turkish speech data.

March 26, 2026Bitlis Eren Üniversitesi Fen Bilimleri Dergisi0 citationsOpen Access

A CNN–NCP Based Hybrid Deep Learning Model for Speech-Driven Gender Classification

SOSevda OlgunNational Yang Ming Chiao Tung University CBCaner BalımAfyon Kocatepe University NONevzat OlgunAfyon Kocatepe University

Key Points

The aim is to investigate a hybrid deep learning model's effectiveness for classifying gender from Turkish speech data.
Developed a hybrid model combining Conv1D layers and neural circuit policies (NCP) for feature extraction.
Utilized mel-frequency cepstral coefficients (MFCC) and log-Mel spectrograms to capture speech properties.
Conducted experiments using a speaker-independent setup with cross-validation on the Mozilla Common Voice Turkish dataset.
Performed ablation studies to assess the impact of NCP components on model performance.
The Conv1D+NCP model achieved 99.29% accuracy and 99.28% F1-score.
The model outperformed conventional LSTM-based models in accuracy and efficiency with fewer parameters.
High computational efficiency indicates suitability for real-time applications.

Abstract

Speech is one of the most natural and effective forms of human communication, carrying both linguistic and non-linguistic information. It plays a crucial role in many applications such as gender classification, biometric authentication, and personalized human-computer interaction. This study aims to investigate the contribution of a hybrid deep learning model based on Neural Circuit Policies (NCP), inspired by biological neural systems, for gender classification on Turkish speech data, by evaluating its performance in terms of accuracy and computational efficiency in comparison with conventional recurrent models. Mel-Frequency Cepstral Coefficients (MFCC) and log-Mel spectrogram features are combined to simultaneously capture the spectral and temporal properties of speech signals. These features are learned as low-level acoustic patterns via Conv1D layers. Long-term temporal dependencies are modeled using Liquid Time Constant (LTC) cells defined within the NCP architecture. To evaluate the generalizability of the model, the experiments were conducted under a speaker-independent setup, and ablation studies were performed by removing different components of the architecture to clearly assess the contribution of the NCP component. Cross-validation was applied on the Mozilla Common Voice 12.0 Turkish dataset during the experiments. The Conv1D+NCP model achieved 99.29% accuracy and 99.28% F1-score, while the LSTM-based model yielded slightly lower results. The NCP-based model offers high performance and computational efficiency with fewer parameters, making it a powerful alternative for real-time applications

AI에게 질문

Bookmark

View Full Paper