Train and deploy a TinyML audio classifier with Edge Impulse

Using Edge Impulse

Now that you’ve learned the foundational concepts of Edge AI and TinyML, it’s time to move from theory to practice.

What is Edge Impulse?

Edge Impulse is a development platform that simplifies building, training, and deploying machine learning (ML) models on embedded systems and edge devices such as microcontrollers, sensors, and single-board computers like the Raspberry Pi or Arduino.

Key Features of Edge Impulse

Why it’s important in Edge AI and TinyML

Edge Impulse offers the following benefits:

• Bridges the gap between machine learning and embedded development
• Accelerates prototyping and deployment of AI features directly on hardware
• Supports TinyML applications that run on devices with very limited memory and compute power
• Works with popular hardware platforms such as Arduino, Raspberry Pi, Nordic, and STMicroelectronics

Get started with Edge Impulse

To begin working with TinyML models, visit Edge Impulse . You’ll need to create a free account to access the full platform.

In the following sections, you’ll walk through each key page on the Edge Impulse platform using snapshots as a guide. These visuals will help you understand what actions to take and how each part of the interface contributes to building and deploying your machine learning model.

Home page of Edge Impulse website

Create a new project

After you create your account and log in, the first step is to create a new project. Give your project a name that clearly reflects its purpose. This helps with easy identification, especially if you plan to build multiple models.

For example, if you’re building a keyword-spotting model, you might name it Wake word detection.

You’ll also need to select the appropriate project type and project settings, as shown in the screenshot below.

New project setup

Configure the target device

After creating your project, the next step is to configure the target device. As you’re using the Arduino Nano RP2040 Connect, click the button highlighted in the screenshot to begin the configuration process. This ensures that data collection, model training, and deployment are optimized for your hardware.

You can find the full specifications for the Arduino Nano RP2040 Connect on Arduino’s official product page .

Follow the settings shown in the screenshot to complete the configuration.

Configure Arduino Nano RP2040

Add the dataset

With your device configured, the next step is to add your dataset to the project. Click Add existing data and follow the configuration settings shown in the screenshot. This lets you upload pre-recorded data instead of collecting it live, which can save time during development.

An example repository has been prepared with assets used throughout this Learning Path. You can clone it with the following command:

    

        
        
git clone https://github.com/e-dudzi/Learning-Path.git

    

The repository contains a Dataset.zip file with the dataset used in this project. Extract the contents to your local machine. For convenience, the dataset is already split into training and testing sets.

Adding existing data

Dataset overview

Dataset uploaded successfully

After the dataset is uploaded, you’ll see the samples listed in the Data acquisition tab, organized by their labels. You can click each sample to inspect the raw signal, view metadata, and even listen to the audio recordings directly within the Edge Impulse interface. This helps confirm that the uploaded data is accurate and ready for training.

Dataset overview

Create the impulse

Now that your data is ready, it’s time to create the impulse. An impulse defines the flow of data from input to output using processing and learning blocks.

Click Create impulse in the menu and configure it as shown in the screenshot. Typically, this includes setting the input data type (such as audio), adding a processing block (like MFCC), and a learning block (such as a neural network classifier).

After configuring the impulse, make sure to save your changes.

Create Impulse

Configure the MFCC block

Next, you’ll configure the MFCC (Mel Frequency Cepstral Coefficients) processing block. This block transforms raw audio into features suitable for training your model.

Click MFCC in the left-hand menu under the Impulse design section.

Set the parameters exactly as shown in the screenshot. These settings determine how the audio input is segmented and analyzed. Once complete, make sure to save the parameters.

These defaults are chosen for this Learning Path, but you can experiment with different values once you’re more familiar with Edge Impulse.

MFCC block configuration

Generate features

After saving the MFCC parameters, the next step is to generate features from your dataset. Click Generate features to begin. Edge Impulse will process your data using the MFCC configuration and create a feature set for training.

When complete, you’ll see a 2D feature plot that shows how the data is distributed across the four labels: on, off, noise, and unknown. This helps visually confirm whether the classes are distinct and learnable.

Feature explorer

Set up the classifier

Now it’s time to configure the neural network classifier, which learns to recognize different audio commands. Click Classifier under Impulse design and set the parameters as shown in the screenshot.

Once all parameters are set, click Save and train to begin training your model.

Review model performance

After training is complete, Edge Impulse displays key performance metrics, including accuracy, loss, and a confusion matrix.

• Accuracy: how often the model correctly predicts labels
• Loss: how far off the predictions are - lower values indicate better performance
• Confusion matrix: shows how well the model classified each label and where it made errors

Review these results to evaluate how well your model learned. If needed, adjust the configuration or retrain to improve accuracy.

On-device performance (EON Compiler – RAM optimized):

You can also use the pre-trained model included in ei-edge-ai-tutorials-arduino-1.0.1.zip from the example repository.

Deploy the model

To run the trained model on your Arduino Nano RP2040 Connect, export it as an Arduino library:

• Click Deployment in the left menu
• Search for “Arduino” and select Arduino library
• If the export doesn’t start automatically, click Build

The model will be downloaded as a .zip file, which you can import into the Arduino IDE.

Next steps

In the next section, you’ll move from model training to real-world deployment. Specifically, you will:

• Connect an LED to the Arduino Nano RP2040 board
• Set up the Arduino IDE for development
• Deploy the trained model to recognize voice commands and control the LED