Research On Obstacle Detection Based On Mobile Phone Ultrasoun

Obstacle detection is very important for the visually impaired to avoid collisions.However,traditional methods usually require carrying additional equipment or are easily affected by light,which does not bring great convenience to people.But ultrasonic waves will not be affected by light,and ultrasonic detection can be realized on low-cost,low-power mobile phones,and it can also achieve good results in real-time detection and long-distance detection.The technology has strong detection ability,and can effectively detect obstacles in indoor open environment,indoor environment with obstacles,outdoor open environment and outdoor environment with obstacles.Ultrasonic waves travel relatively fast through most materials,so they can be detected without damaging surrounding objects.This paper proposes a system that uses mobile phone ultrasound to detect whether there is an obstacle,which is a technology based on the built-in sensor of the mobile phone.By transmitting ultrasonic signals and receiving and analyzing the difference of the reflected acoustic signals,it can detect whether there are obstacles in the environment.By studying the transmitted and received acoustic signals and extracting features,the information of obstacles can be detected,thereby helping the visually impaired to travel.The research content of this paper for mobile phone ultrasonic obstacle detection is as follows:1.Modulate FM CW(Frequency Modulated Continuous Wave,FMCW)and sine wave.Use MATLAB to simulate the radar system,and transmit and receive FMCW.The distance between the mobile phone and the obstacle causes the frequency of the reflected signal to change.Calculate the frequency difference between the transmitted signal and the reflected signal to calculate the distance between the mobile phone and the obstacle.Modulate the sine wave,the speaker of the mobile phone sends out the sine wave and the microphone of the mobile phone receives the recorded echo at the same time,and detects the obstacle information by analyzing the frequency change.Select the sine wave to detect obstacles by comparison.In order to get more useful signals,this paper also denoises the obtained signals through band-pass filtering.2.Feature selection of sound waves.In this paper,multiple sets of audio data collected are pre-filtered and framed to obtain Mel Frequency Cepstral Coefficients(MFCCs).Connect the data sampling points to the data to obtain the time domain profile.Obtain Doppler shift using STFT.The collected three kinds of small data features are trained and classified respectively,and the convolutional neural network and machine learning training3.This paper designed an obstacle detection system based on mobile phone ultrasonic waves model are used to finally select the feature Doppler frequency shift with the best obstacle detection effect.3.This paper designs an obstacle detection system based on mobile phone ultrasound to help the visually impaired travel.In this paper,the transmitted sine wave is modulated to detect obstacles.In this paper,a band-pass filter is used to filter the collected sound waves,and the Doppler frequency shift feature is obtained through Short Time Fourier Transform(STFT).This paper uses Generative Adversarial Networks(GAN)to Augment the dataset and feed the raw and generated data into the neural network model for classification recognition.This paper builds a lightweight Convolutional Neural Network(CNN)training model and compares a variety of other machine learning models.The results show that the lightweight CNN network model built in this paper is better than random forests,decision trees,etc.The final accuracy rate can reach more than 93%.