This study converts piano recordings to mel spectrogram and classifies them by SOTA pre-trained neural network backbones in CV. Comparative experiments show that SqueezeNet achieves a best classification accuracy of 92.37%.|该项目将钢琴录音转为为mel频谱图，使用微调后的前沿计算机视觉领域预训练深度学习骨干网络对其进行分类，对比实验可知SqueezeNet作为最优网络正确率可达92.37%

Golang Mel Spectrogram and Spectrogram inversion

This repository contains the code and methodology used for the BirdCLEF 2024 Kaggle competition, where I achieved a rank of 55th out of 974 participants, earning a bronze medal. The goal of this competition was to build a model that can accurately classify bird sounds.

Project to classify wav audio files using a CNN.

Speech Emotion Recognition (SER) using CNNs and CRNNs Based on Mel Spectrograms and Mel Frequency Cepstral Coefficients (MFCCs)

This Model analyzes and predicts the input sound and then using pretrained ANC systems cancels the input sound.

In this project we used TESS voice dataset and processed it and perform emotion prediction.

End-2-end speech synthesis with recurrent neural networks

Zafar's Audio Functions in Julia for audio signal analysis: STFT, inverse STFT, CQT kernel, CQT spectrogram, CQT chromagram, MFCC, DCT, DST, MDCT, inverse MDCT.

Zafar's Audio Functions in Matlab for audio signal analysis: STFT, inverse STFT, mel filterbank, mel spectrogram, MFCC, CQT kernel, CQT spectrogram, CQT chromagram, DCT, DST, MDCT, inverse MDCT.

Zafar's Audio Functions in Python for audio signal analysis: STFT, inverse STFT, mel filterbank, mel spectrogram, MFCC, CQT kernel, CQT spectrogram, CQT chromagram, DCT, DST, MDCT, inverse MDCT.

Easier audio-based machine learning with TensorFlow.

Deliverables relating to the Speech Technology University Unit (Notes Courtesy to Dr. Andrea De Marco)

Open Source Repository for the MASA Project

Step onto the stage with Saxophone Hero, where your tenor saxophone is the key to unlocking a rhythmic adventure through a world of sheet music. In this game, your character scores points by hitting the right notes. Powered by machine learning, the game captures the pitch from your saxophone and translates it to player movement in real time.

Embed chiptunes in 2D with Convolutional Auto Encoder and Mel Spectrograms

Least-squares (sparse) spectral estimation and (sparse) LPV spectral decomposition.

Different Signal Processing Tasks

Leveraged Dynamic Time Warping (DTW) to assess the similarity between specific audio tracks

Created an ASR (Automatic Speech Recognition) system that takes in individual recordings. Each recording represents a sentence composed of 5-10 English language digits, separated by adequate pauses. The system involves segmenting the sentence using a classifier, differentiating between background and foreground sounds.