Audio Classification System Using Neural Network for Vocal Percussion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems lack effective methods to classify and recognize vocal percussion sounds, such as beatboxing, which mimic musical instruments, into specific drum sounds like bass drum or snare drum, limiting the ability to generate corresponding musical notes or patterns.
Innovation Solution
A system utilizing a trained convolutional neural network with dropout layers and Frequency Cepstral Coefficient (MFCC) feature extraction layers to classify audio inputs from microphones, transforming audio inputs into images for accurate percussion sound recognition, enabling the generation of digital data for musical instruments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a trained neural network with MFCC feature extraction and image transformation is used to classify vocal percussion sounds, then the classification accuracy is improved, but the device complexity increases
Solution Approach 1:
The patent introduces MFCC feature extraction as an intermediary processing step that transforms raw audio signals into meaningful feature representations. This intermediary layer bridges the gap between complex audio input and the neural network classifier, enabling accurate vocal percussion classification while managing system complexity through specialized feature engineering
Solution Approach 2:
The patent replaces traditional audio signal processing methods with a neural network-based classification system. The mechanical/audio processing approach is substituted with computational intelligence, using deep learning models to automatically learn and classify vocal percussion patterns from audio inputs
2Measurement precision
If audio input is transformed into images using MFCC feature extraction for classification, then the recognition capability is improved, but the processing time increases
Solution Approach 1:
The patent applies MFCC feature extraction as a preliminary action before the main classification task. By pre-processing audio inputs into compact feature representations and transforming them into image formats, the system prepares optimized data structures that accelerate subsequent neural network processing and classification operations
3Measurement precision
If multiple input channels (kick, snare, etc.) are processed separately through the neural network, then the classification precision is improved, but the device complexity increases
Solution Approach 1:
The patent segments the audio classification task into separate processing channels for different percussion types (kick, snare, etc.). Each input channel is processed independently through the neural network, allowing specialized feature extraction and classification for each percussion instrument type, thereby improving overall classification precision
Data Source
AI summary
A system includes a computer including a processor and a memory. The memory includes instructions such that the processor is programmed to receive an audio input representing a percussion performed by a user and classify, at a trained neural network, the audio input as a particular musical type.