Ultrasonic Sensor Data for Human or Nonhuman Classification

Title:Ultrasonic Sensor Data for Human or Nonhuman Classification

Author:Budi SetiadiPoliteknik Negeri Bandung, IndonesiaEmail: [[e-mail]]

Year: 2025

Description:The dataset consists of distance measurements from 11 ultrasonic sensors (HC-SR04) arranged vertically on a wall with 10 cm spacing and positioned between 80 cm and 180 cm above the floor. Each sensor measures the reflected distance of passing objects to recognize their vertical distance patterns.The dataset includes two object categories:1.    Human – visually impaired individuals.2.    Nonhuman – inanimate objects such as chairs or other static items.

Keywords:Ultrasonic Sensor Array, MLP, Classification.

File consist of three:- MLP_Ultrasonic_Classification_Human-Nonhuman_Training1400.csv- MLP_Ultrasonic_Classification_Human-Nonhuman_Testing600.csv- README_MLP_Ultrasonic_Classification_Human-Nonhuman_Dataset_2025.txt 

Data Acquisition ProcessThe data were collected directly at SLBN-A Citeureup, Cimahi City, using an ultrasonic sensor array connected to a microcontroller through multiplexers (CD74HC4067). A total of 10 human subjects participated in repeated testing sessions to produce 2,000 ultrasonic sensor readings representing both human and nonhuman reflection patterns.Each measurement contains 11 distance values (in centimeters) corresponding to sensors 1–11, and one label indicating the object class (“human” or “nonhuman”). All data were stored in .csv format.

Dataset PurposeThe dataset aims to:•    Train and evaluate Multi Layer Perceptron (MLP) for object classification based on ultrasonic distance patterns.•    Provide representative data reflecting vertical ultrasonic reflection patterns between humans and nonhuman objects.•    Support the development of adaptive navigation assistance systems for visually impaired individuals.

Data StructureColumn    Description    Data TypeSensor₁ … Sensor₁₁    Distance values from each of the 11 ultrasonic sensors (in cm)    Numeric (float/int)Label    Object class (“Human” / “Nonhuman”)    Categorical (binary)

The dataset was divided into two subsets:•    Training set (70%) → 1,400 samples.•    Testing set (30%)  600 samplesusing a stratified sampling method to maintain class balance

License:Creative Commons Attribution 4.0 International (CC BY 4.0)https://creativecommons.org/licenses/by/4.0/

Suggested Citation:Budi Setiadi (2025). Ultrasonic Sensor Data for Human or Nonhuman Classification. Zenodo. DOI: [10.5281/zenodo.17570045]

Summary•    Total samples: 2,000•    Number of features: 11 input sensors + 1 output label•    Acquisition site: SLBN-A Citeureup, Cimahi City, Indonesia•    File format: .csv•    Purpose: Training and testing of a MLP classification model for adaptive object detection and navigation aid for visually impaired individuals.