Speech Emotion Recognition using Machine Learning and Librosa

Aug 2, 2026

Upload by Riley

Speech emotion recognition plays a key role in human-computer interaction by enabling systems to infer and respond to human affect. This paper presents a machine-learning based approach for classifying emotions from speech signals using librosa for audio feature extraction. Multiple classifiers, including Support Vector Machine, Random Forest, and k-Nearest Neighbors, are trained to identify emotions such as happy, sad, angry, neutral, and surprised from speech audio.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages7

Reading Time18 min

AccessFree to download

I. Introduction

Dataset Description

II. Methodology

Data Preprocessing

Machine Learning Models

Model Training and Hyperparameter Tuning

Model Evaluation

International Journal of Advanced Research in Education and TechnologY (IJARETY) Volume 12, Issue 1, January-February 2025 Impact Factor: 7.394 [www.ijarety.in](www.ijarety.in) editor.ijarety@gmail.com International Journal of Advanced Research in Education and TechnologY(IJARETY) | ISSN: [2394-2975 | ](2394-2975 | www.ijarety.in| | Impact Factor:)[www.ijarety.in](2394-2975 | www.ijarety.in| | Impact Factor:)[| | Impact Factor:](2394-2975 | www.ijarety.in| | Impact Factor:) 7.394| A Bi-Monthly, Double-Blind Peer Reviewed & Referred Journal | || Volume 12, Issue 1, January-February 2025 || DOI:10.15680/IJARETY.2025.1201031 Speech Emotion Recognition using Machine Learning and Librosa Pavithra J, Sivashree S Department of Computer Science and Engineering, Jeppiaar SRR Engineering College, Chennai, Tamil Nadu, India ABSTRACT: Emotion recognition from speech is an important aspect of human-computer interaction (HCI) systems, allowing machines to better understand human emotions and respond accordingly. This paper explores the use of machine learning techniques to recognize emotions in speech signals. We leverage the librosa library for feature extraction from audio files and train multiple machine learning models, including Support Vector Machine (SVM), Random Forest (RF), and k-Nearest Neighbors (k-NN), to classify speech emotions. The aim is to create an automated system capable of identifying emotions like happy, sad, angry, neutral, and surprised from speech audio. I. INTRODUCTION Emotion recognition from speech is a challenging task, as human emotions are often subtle and context-dependent. However, with advancements in machine learning and audio signal processing, it is now possible to analyze speech data and classify the underlying emotions. The process typically involves extracting audio features such as Melfrequency cepstral coefficients (MFCCs) , spectral features, and prosodic features, which are then used as input to machine learning models for classification. In this study, we focus on building an emotion recognition system using the librosa library, which is widely used for audio analysis in Python. We then apply machine learning techniques, including SVM, Random Forest, and k-NN, to classify speech audio files into various emotion categories. Dataset Description For this experiment, we use the RAVDESS (Ryerson Audio-Visual Database of Emotional Speech and Song) dataset, which is a well-known dataset for emotion recognition. It contains 1,440 audio files from 24 professional actors expressing 8 different emotions: neutral, calm, happy, sad, angry, fearful, disgust, and surprised. For simplicity, this paper focuses on a subset of five emotions: happy, sad, angry, neutral, and surprised. Each file in the dataset contains a 3-second audio clip, and the emotion is labeled with a numerical identifier. The task is to classify each audio clip into one of the five emotion categories. II. METHODOLOGY Data Preprocessing 1. Loading Audio Files: The audio files are loaded using the librosa library, which provides efficient methods for reading and processing audio. 2. Feature Extraction: We extract several features from the speech audio using librosa: o Mel-frequency Cepstral Coefficients (MFCCs): These coefficients represent the short-term power spectrum of the sound and are commonly used in speech recognition and emotion classification. o Chroma Features: These describe the harmonic content and are particularly useful in music and speech analysis. o Spectral Contrast: This feature measures the difference in amplitude between peaks and valleys in a sound spectrum. o Zero-Crossing Rate: The rate at which the signal changes sign, often associated with the noisiness of a sound. o Spectral Roll-off: The frequency below which a certain percentage of the total spectral energy lies. The extracted features are used to represent the emotional content of the speech. 3. Data Normal

FAQs

What is the main goal of this paper?

The paper aims to build an automated system that recognizes emotions from speech audio by classifying clips into categories such as happy, sad, angry, neutral, and surprised.

Which audio features are extracted using librosa?

It extracts MFCCs, chroma features, spectral contrast, zero-crossing rate, and spectral roll-off, then uses the resulting features to represent emotional content.

Which machine learning models are used for emotion classification?

Support Vector Machine, Random Forest, and k-Nearest Neighbors are applied to the preprocessed feature data for classification.

Aug 2, 2026

Upload by Riley