This section contains Research and Development projects in Machine Learning and Deep Learning that require original developments. They call on our expertise in Digital Signal Processing, Optimization, Calculus, Linear Algebra.
Automatic environmental sound classification (ESC) based on ESC-50 dataset (and ESC-10 subset) built by Karol Piczak and described in the following article:
"ESC: Dataset for Environmental Sound Classification." by Karol J. Piczak. 2015. In Proceedings of the 23rd ACM international conference on Multimedia (MM '15). Association for Computing Machinery, New York, NY, USA, 1015–1018. https://doi.org/10.1145/2733373.2806390"
Multi-feature Convolutional Neural Networks (CNN) achieves accuracy close to 99%, with custom pre-processing and a fusion mel-spectrograms + complex wavelet transforms.
The last confusion "sea waves" "rain" is solved by developping an original transform of the complex CWT. This Transform, aT-CWT replaces the phase of the CWT for stationary, pseudo-stationary sounds with a Gaussian distribution.
With the aT-CWT transform, the multi-feature CNN model achieves 100% accuracy.
- Concentrating on a specific machine type: valves.
- Denoising the recordings using MVDR beamforming combined with a custom, fixed Generalized Sidelobe Canceler (GSC).
- Applying unsupervised classification techniques (auto-encoder, etc.) to two sets of signals: single microphone recordings and the denoised GSC output, to detect defective valves and demonstrate the benefits of MVDR beamforming combined with GSC.
We develop an automatic unsupervised classification model or automatic diagnosis model for detecting failures or breakdowns of industrial machinery based on their acoustics characteristics, recorded with a 8-microphones circular array.
The model is based on the MIMII dataset made available by Hitachi, Ltd. under a Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license.
https://zenodo.org/records/3384388
Unlike most classification models found in literature, this study somewhat deviates from the initial challenge's rules: classification of noisy signals. However, since we have access to multiple channels, it makes practical sense to denoise the signals before initiating the classification process. Thus, the challenge here is transforming the 8-microphone array into a "sensor" for monitoring industrial machinery sounds in noisy environments. Then, we apply the classification model to these denoised signals to automatically identify anomalies, failures, or breakdowns.
Rather than classifying various types of machines (pumps, fans, valves, sliders), our focus will be:
Applications
- Rotating machinery Failure Detection: bearings, motors,rotors.
- HVAC Fault detection and diagnosis (FDD): pumps, compressors, valves.
In this project we develop effective methods for classifying mitochondrial genomes (DNA sequences) based on Digital Signal Processing, Machine Learning, Deep Learning. This is on-going research and results will be published on a regular basis. As a starting point we analyzed the following paper:
"ML-DSP: Machine Learning with Digital Signal Processing for ultrafast, accurate, and scalable genome classification at all taxonomic levels" by Gurjit S. Randhawa , Kathleen A. Hill and Lila Kari. https://doi.org/10.1186/s12864-019-5571-y
Their alignement free DNA sequence classification approach: ML-DSP is very effective.
By introducing a simple alignment technique and short FFTs: ML-FFT + SoftAlign, we outperform ML-DSP with difficult datasets: Fungi, Insects.
This section is a portfolio of Machine Learning projects with Python and various visualization and analysis tools. Most of these projects were carried out within the framework of IBM certifications. They are presented with Jupyter Notebooks.
Some projects have been improved by incorporating more in-depth data analysis, better graphs, advanced ML techniques.
-
In this project, we predict if the Falcon 9 first stage will land successfully. Project includes: SpaceX data collection, Data Wrangling, Webscraping, EDA with SQL Queries & Data visualization, SpaceX Launch Records Dashboard, Launch Sites Locations Analysis with Folium, Machine Learning classification with optimization of hyperparameters and selection of best model: KNN, Decision Tree, SVM, Logistic Regression.
-
A widerange of small projects with various ML techniques, prediction, supervised and unsupervised classification: Linear Regression, Polynomial Regression, Non-Linear Regression, Recommandation Systems, KNN, Customer Segmentation with K-Means, Hierarchical Clustering, Density-Based Clustering, Logistic Regression.
-
The project consists of finding the best model for predicting home prices in King County, USA in Washington State, based on a dataset of homes sold between May 2014 and May 2015. Prediction accuracy was improved by implementing a spline regression model.
One Jupyter Notebook includes interactive Folium maps (interactive maps will not display on Github).
-
Old dataset on housing prices derived from the U.S. Census Service to present insights based on our experience in Statistics. Median value of houses bounded by the Charles river, of owner-occupied units built before 1940, relationship between Nitric oxide concentrations and the proportion of non-retail business acres per town, impact of weighted distance to the five Boston employment centres on the median value of owner-occupied homes.
🔭 I’m currently working on advanced projects in ML & DL
👯 I’m looking to collaborate on Digital Signal Processing, Machine Learning, Deep Learning
📫 How to reach me: stephane.dedieu@bloo-audio.com