SDAA429A June   2026  – October 2026 MSPM0G5187

 

  1.   1
  2.   Abstract
  3.   Trademarks
  4. 1 Introduction
  5. 2 MSPM0G5187 with TinyEngine NPU
  6. 3 Edge AI Toolchains
    1. 3.1 TI Edge AI Studio
    2. 3.2 TI Tiny ML Tensorlab
    3. 3.3 TI Neural Network Compiler
  7. 4 Rapid Prototyping Through TI Edge AI Studio
    1. 4.1 Project Setup
    2. 4.2 Data Capture
      1. 4.2.1 Hardware Setup
      2. 4.2.2 Feature Overview
    3. 4.3 Data Import
      1. 4.3.1 Classification Task
      2. 4.3.2 Regression Task
      3. 4.3.3 Forecasting Task
      4. 4.3.4 Annotation List
    4. 4.4 Data Annotation
    5. 4.5 Model Training
      1. 4.5.1 Training Parameter
      2. 4.5.2 Toolchain Dependency
    6. 4.6 Performance Reports
    7. 4.7 Model Export
  8. 5 Advanced Development Through TI Tiny ML Tensorlab
    1. 5.1 Environment Setup
    2. 5.2 Data Acquisition
      1. 5.2.1 Data Retrieval
      2. 5.2.2 Data Validation
    3. 5.3 Model Architecture Iterative Refactoring
      1. 5.3.1 Baseline Implementation
      2. 5.3.2 Refactoring With PyTorch Built-in Modules
      3. 5.3.3 Data Pipeline Optimization
      4. 5.3.4 Transition to CNN Architecture
    4. 5.4 Model Export
    5. 5.5 Quantization-Aware Training
    6. 5.6 Model Compilation
  9. 6 Edge AI Application: Digit Recognition
    1. 6.1 LeNet-5 Variant CNN Model
    2. 6.2 NPU/CPU Performance Comparison
  10. 7 Edge AI Application: Waveform Classifier
    1. 7.1 Feature Extraction
    2. 7.2 Time-Series Classification Model
    3. 7.3 Model Memory Considerations
    4. 7.4 NPU/CPU Performance Comparison
  11. 8 Summary
  12. 9 References
  13. 10Revision History

Feature Extraction

Feature extraction is a critical component of common EdgeAI applications, particularly for audio and time-series signal processing. The feature extraction module provides efficient signal processing functions optimized for embedded systems, leveraging ARM libraries for optimal performance. The available feature extraction functions include:

  • Real Fast Fourier Transform (RFFT): Converts time-domain signals to frequency domain representation
  • Magnitude Scaling: Applies scaling for normalizing signal magnitude to appropriate range and prevent overflow/underflow for subsequent processing
  • Complex Magnitude Calculation: Calculates the magnitude of complex numbers resulting from FFT
  • DC Component Removal: Eliminates signal DC bias for improved feature accuracy
  • Bin Calculation: Averages frequency components within defined bins for dimensionality reduction

Figure 7-1 shows the feature extraction processing pipeline.

 Feature Extraction Processing PipelineFigure 7-1 Feature Extraction Processing Pipeline

Table 7-6 shows the feature extraction processing pipeline configuration designed for waveform classifier. When the corresponding configuration symbol is defined, its associated task is executed within the processing pipeline.

Table 7-3 Pipeline Configuration for Waveform Classifier
Processing TaskConfiguration SymbolDefined or Not
RFFTFE_RFFTY
Magnitude ScalingFE_COMPLEX_MAG_SCALEY
DC Component RemovalFE_DC_REMY
Complex Magnitude CalculationFE_MAGY
Bin CalculationFE_BINY

Table 7-4 shows the feature extraction configuration parameters designed for waveform classifier. For each inference task, eight frames are concatenated, where each frame processes 256 data points to generate 16 feature dimensions, resulting in a total of 128 feature dimensions fed into the waveform classifier AI model.

Table 7-4 Waveform Classifier Feature Extraction Parameters
MetricConfiguration SymbolValue
Input Frame SizeFE_FRAME_SIZE256
Feature Size per FrameFE_FEATURE_SIZE_PER_FRAM16
Frequency Bin SizeFE_BIN_SIZE8
Normalize Bin ValuesFE_BIN_NORMALIZE1
Frame to ConcatenateFE_NUM_FRAME_CONCAT8
Input Feature SizeFE_STACKING_FRAME_WIDTH128
Note: The feature extraction parameters configured in Table 7-4 must match the parameters used during model training to verify consistent results.

Performance Consideration

  • Optimized Computation: Leverages ARM-optimized libraries for efficient processing
  • Memory Efficiency: Buffer swapping minimizes memory copy overhead
  • Dimensionality Reduction: Bin calculation reduces feature dimensions, lowering memory usage and computation in downstream stages
  • Parameters Tuning: Configuration parameters must be tuned to match available memory and compute resources