Signal and Image processing (07_O-TSI)

• Coefficient : 6
• Hourly Volume: 150h (including 72h supervised)
  CM : 34.5h supervised
  TD : 4.5h supervised
  Labo : 33h supervised (and 12h unsupervised)
  Out-of-schedule personal work : 66h
• Including project : 3h supervised and 10h unsupervised project

AATs Lists

Description

1. Digital signal processing
   • Actual sampling and restitution
   • Synthesis of analog filters
   • Synthesis of digital filters
   • Spectral analysis of modulated signals
   • Analog and digital correlation
2. Digital image processing
   • Acquisition and representation of images
   • Restoration and pre-treatments
   • Segmentation by contours
   • Segmentation by regions
   • Classification and pattern recognition
3. DSP: Interest, architecture and implementation
   • DSP applications and performance
   • Architecture, instructions and specific addressing modes, peripherals
   • Assembler and C programming, addressing modes
   • Execution time (Benchmarking)
   • Generating DSP code from MATLAB-Simulink
4. DSP and Matlab: Application examples
   • Digital audio filtering (FIR, IIR)
   • FFT for spectral analysis
   • Coding sound effects
   • Image denoising
   • Image segmentation

Learning Outcomes AAv (AAv)

• AAv1 [heures: 12, B2, B3, B4]: At the end of the semester, the student must be able to do the modeling and spectral analysis of real sampling mechanisms, in particular the averaging sampler (or tracker ) and the hold sampler (or hold). The student must know the influence of the choice between these two processes on the spectrum of the sampled signal and take it into account when designing the two filters, the guard filter (upstream of sampling) and the reconstruction filter or interpolator.

• AAv2 [heures: 6, B2, B3, B4, C2]: At the end of the semester, the student will be able to determine the correlation functions (intercorrelation, autocorrelation) and the spectral energy density (DSE) or power (DSP ) of deterministic signals. It must also be able to apply correlation in radar detection to detect the presence of a pattern in a received signal.

• AAv3 [heures: 12, B2, B3, B4, C1, C2, D1, D3]: At the end of the semester, the student must be able to do the modeling and spectral analysis of the principles of modulation and demodulation of amplitude and frequency. The student must know how to analyze and interpret the temporal and frequency representations of analog signals corresponding to the following modulation formats: AM (dual band with DSB carrier, dual band with suppressed carrier DSB-SC, single sideband SSB) and FM (narrow band , wideband). He must also know how to use simulation tools (python, matlab or octave) and the spectrum analyzer to carry out demodulation by envelope detector or synchronous detector.

• AAv4 [heures: 28, B2, B3, B4, C1, C2, D1, D2, D3] : At the end of the semester, the student must be able to design, analyze and implement digital filters of type RII or RIF in response to specifications in a specification. To successfully complete this work, the student must be able to: (1) Translate the specifications into a template. (2) Appropriately choose a filter structure (RII or RIF) and a synthesis method (bilinear transformation, impulse invariance or transfer function sampling) by arguing the relevance of the choices made. (3) Determine the filter coefficients by direct calculation or using a matlab/simulink type rapid prototyping tool. (4) Implement the filter in an interpreted language such as Python, Matlab or Octave and validate its performance against the specified template. He must also be able to study the influence of the frequency distortion implied by the synthesis method. (5) Choose a form (direct, cascade or parallel) of implementation. It must also be able to study the influence of the frequency distortion implied by the quantification of the filter on a finite number of bits (sensitivity to the finite representation of the coefficients). (6) Implement the filter on a microcontroller or DSP type hardware target. (7) Validate the synthesis against the specifications by measurement using a spectrum analyzer.

• AAv5 [heures: 21, C1, C2, D1, D2, D3] : At the end of the semester, the student must be able to design, analyze and implement a digital synthesizer with subtractive synthesis supporting the MIDI communication protocol ( Musical Instrument Digital Interface) dedicated to music. To successfully complete this work, the student must be able to: (1) Generate basic sound signals such as sine, square, triangle, sawtooth by table reading. The frequency of these signals must be a function of the note entered on the MIDI keyboard. The amplitude must be modulated over time by an ADSR type envelope (Attack Decay Sustain Release for Attack Decay Maintenance Extinction in French). (2) Simulate and implement RII or RIF type digital filtering whose resonance and cutoff frequency are adapted to the note received. Amplitude Envelope Management (ADSR) should bring the generated sound to life. (3) Add digital sound processing to generate Reverb (reverberation) or polyphony type effects. (4) Implement these sound synthesis algorithms on a microcontroller or DSP type hardware target.

• AAv6 [heures: 40, B2, B3, B4, C2,] : At the end of the semester, the student will be familiar with the challenges of artificial vision and will have acquired the fundamental concepts of processing and analysis of digital images 2D. This concerns: (1) the representation of images in the spatial and frequency domain, (2) contrast improvement by histogram modification techniques (linear and non-linear anamorphosis), (3) denoising by techniques linear filtering (2D convolution operators) and non-linear (filtering of order statistics, image averaging, morphological transformations), (4) restoration by contrast enhancement operations (deblurring), (5) segmentation by contour-based and region-based approaches, (6) texture analysis by frequency and statistical approaches, (7) feature extraction using attribute selection tools, (8) recognition of objects by Hough transform and by machine learning algorithms.

• AAv7 [heures: 15, C2, C3, D1, D2, D3, D4]: At the end of the semester, the student will be able to effectively apply one or more classic processing and image algorithms to an input image. image analysis. He must be able to optimize the parameterization of each algorithm and analyze the relevance and limits of the results obtained.

• AAv8 [heures: 12 , C1, C2, C3, D1, D2, D3, D4] : At the end of the semester, the student will be able to design, analyze and implement a processing and analysis chain of images in response to specifications reflecting the needs of a new computer vision application. This involves in particular: (1) finding the right preprocessing operator with regard to the nature of the noise in the image (Gaussian, impulsive or uniform), (2) making a justified choice on the method and on the segmentation operator to use, (3) know how to identify the right characteristic attributes for the analysis and exploitation of the information present in the image, (4) choose an object recognition algorithm adapted to the problematic, (5) implement the algorithms in an interpreted language such as matlab or octave and finally (6) carry out the necessary tests to validate the proposed solution and critically evaluate the results obtained.

• AAv9 [heures: 6, D1, D2, D3, D4]: At the end of the semester, the student will be able to use the tools of the openCV library and implement the implementation of a processing and analysis solution images on a microcontroller type card connected to a camera.

• AAv10 [heures: 6, D2] : At the end of the semester, the student will be able to use the libraries of deep learning techniques.

Assessment methods

Average of several assessments

Key Words

filter synthesis, modulated signals, correlation, image denoising, image segmentation, pattern recognition, DSP

Prerequisites

Signal processing and mathematics curriculum from previous years.

Resources

Course handouts and tutorial and lab texts