Machine Learning and Deep Learning with Wavelets
Derive low-variance features from real-valued time series and image data for use in machine learning and deep learning for classification and regression. Use continuous wavelet analysis to generate the 2D time-frequency maps of time series data, which can be used as inputs with deep convolutional neural networks (CNN).
Analyze signals jointly in time and frequency and images jointly in space, spatial frequency, and angle with the continuous wavelet transform (CWT).Use wavelet coherence to reveal common time-varying patterns. Perform adaptive time-frequency analysis using nonstationary Gabor frames with the constant-Q transform (CQT).
Discrete Multiresolution Analysis
Perform decimated discrete wavelet transform (DWT) to analyze signals, images, and 3D Volumes in progressively finer octave bands. Implement nondecimated wavelet transforms. Decompose nonlinear or nonstationary processes into intrinsic modes of oscillation using techniques.
Use orthogonal wavelet filter banks like Daubechies, Coiflet, Haar and others to perform multiresolution analysis and feature detection. Design custom filter banks using the lifting method. Lifting also provides a computationally efficient approach for analyzing signal and images at different resolutions or scales.
Denoising and Compression
Use wavelet and wavelet packet denoising techniques to retain features that are removed or smoothed by other denoising techniques. The Wavelet Signal Denoiser app lets you visualize and denoise 1D signals. Use wavelet and wavelet packets to compress signals and images by removing data without affecting perceptual quality.
Acceleration and Deployment
Speed up your code by using GPU and multicore processors for supported functions. Use MATLAB Coder™ to generate standalone ANSI-compliant C/C++ code from Wavelet Toolbox functions that have been enabled to support C/C++ code generation. Generate optimized CUDA code to run on NVIDIA® GPUs for supported functions.