PERANCANGAN DAN IMPLEMENTASI EQUALIZER 20 CHANNELS BERBASIS SEMANTIC MENGGUNAKAN NATURAL LANGUAGE PROCESSING DAN MACHINE LEARNING PADA PLATFORM ANDROID

- Mohamad Fikri Haikal Anwar, Arjuni Budi Pantjawati¹

1. ¹ Universitas Pendidikan Indonesia

Repository at Universitas Pendidikan Indonesia (Universitas Pendidikan Indonesia), 2026

Meningkatnya kebutuhan akan pengolahan audio yang intuitif pada perangkat mobile mendorong lahirnya sistem equalizer berbasis semantik sebagai solusi cerdas dalam kendali suara. Penelitian ini bertujuan merancang dan mengimplementasikan equalizer 20 channels berbasis semantic control menggunakan teknologi Natural Language Processing (NLP) dan Machine Learning (ML) pada platform Android. Sistem memungkinkan pengguna memberikan perintah berupa teks atau suara, yang diproses oleh NLP berbasis FastText menjadi representasi vektor. Vektor ini kemudian dipetakan ke konfigurasi gain equalizer melalui model Multi-Layer Perceptron (MLP) dengan pendekatan supervised learning. Model dilatih menggunakan pasangan deskripsi semantik dan nilai gain, dioptimalkan dengan Adam Optimizer serta early stopping, lalu dikonversi ke format TensorFlow Lite agar efisien di Android. Integrasi sistem dilakukan pada aplikasi berbasis Android Studio dengan antarmuka interaktif yang mendukung input audio, perintah semantik, serta visualisasi gain pada 20 channels. Hasil pengujian menunjukkan sistem mampu memetakan perintah semantik dengan akurasi tinggi, menjaga konsistensi frequency response, serta beroperasi secara real-time dengan latensi rendah. Dengan demikian, penelitian ini berhasil menghadirkan equalizer berbasis semantik yang cerdas, intuitif, dan responsif pada perangkat Android. The increasing demand for intuitive audio processing on mobile devices has driven the development of semantic-based equalizer systems as an intelligent solution for sound control. This study aims to design and implement a 20-channel equalizer with semantic control using Natural Language Processing (NLP) and Machine Learning (ML) technologies on the Android platform. The system allows users to provide commands in the form of text or speech, which are processed by FastText-based NLP into vector representations. These vectors are then mapped to equalizer gain configurations through a Multi-Layer Perceptron (MLP) model using a supervised learning approach. The model is trained with pairs of semantic descriptions and gain values, optimized with the Adam Optimizer and early stopping, and converted into TensorFlow Lite format for efficiency on Android devices. The system integration is carried out within an Android Studio-based application, featuring an interactive interface that supports audio input, semantic commands, and visualization of gain across 20 channels. The experimental results show that the system is capable of mapping semantic commands with high accuracy, maintaining consistent frequency response, and operating in real-time with low latency. Thus, this study successfully presents a semantic-based equalizer that is intelligent, intuitive, and responsive on Android devices.