Newborn crying sound recognizing method based on dynamic time warping

Abstract

The invention belongs to the technical field of biological feature recognition, and specifically provides a newborn crying sound recognizing method based on dynamic time warping. The method comprisesthe following steps: collecting newborn crying sounds under three conditions, and marking the newborn crying sounds as the crying sounds caused by hunger, the crying sounds caused by pain and the crying sounds caused by unknown reasons; carrying out preprocessing including downsampling, pre-emphasis, framing windowing, endpoint detection and the like on the crying sounds, and carrying out featureextraction on the preprocessed crying sound signals, namely, extracting the short-time energy, the fundamental tone frequency and the 0-12-level mel frequency cepstrum coefficient; carrying out one-way variance analysis on the feature parameters, and screening out the features with significant differences from the three types of crying sounds; selecting reference templates of three types by adopting the dynamitic time warping algorithm; and carrying out three two-class tasks, carrying out matching recognition by adopting the DTW algorithm, and calculating the recognition rate. For the method,the computation complexity is relatively low, the alkylating speed is accelerated, in addition, in the dynamic time warping algorithm, the equal length of two samples is not needed, and thus the earlier-stage processing step is saved.

Description

technical field [0001] The invention belongs to the technical field of biological feature recognition, and in particular relates to a newborn cry recognition method. Background technique [0002] Newborns do not have the ability to express language, and the only way to express their needs and emotions is to cry. Excessive crying of newborns is the most common complaint of mothers, making them feel anxious and overwhelmed, which is also an important reason for postpartum depression. Excessive crying and noise also brings a lot of workload and confusion to clinical departments and pediatric care departments. [0003] At present, most of the research on newborn crying focuses on the analysis of the difference in crying in different states, mainly starting from visual differences such as spectrograms. The purpose of classification and recognition is to compare the recognition rates of different features. Most of them use classifiers for identification, which requires a large d...

AI Technical Summary

Problems solved by technology

Most of them use classifiers for identification, which requires a large data set and occupies a large computer memory

Images