Infant cry detection deep learning method

Abstract

The invention discloses an infant cry detection deep learning method, and relates to the technical field of voice signal processing. The method comprises the following steps: a, acquiring a voice signal; b, framing the voice signal segment and performing cochlear voice feature extraction on each frame; and c, inputting the adjacent N frames of voice features into a pre-trained baby cry detection deep learning model for reasoning and judging whether cry exists or not and d, voting the N frames of classification results by applying a majority-first voting principle, and judging whether baby cryexists in the N frames or not. The cochlear speech features adopted by the method are speech feature parameters more conforming to auditory perception analysis of people, and the convolutional networkand the long-short-term memory recurrent neural network are adopted as acoustic inference models for infant cry detection, so that the method can adapt to a speech environment with a low signal-to-noise ratio, and has higher accuracy compared with a traditional method.

Description

technical field [0001] The invention belongs to the technical field of voice signal processing, and in particular relates to a deep learning method for baby crying detection. Background technique [0002] In modern society, parents or elders tend to neglect newborn babies due to busy work or housework, and babies can only express their emotions and needs by crying. Therefore, there is a great market demand for home care based on baby cry detection. [0003] The current speech signal recognition system usually consists of three parts, which are speech signal preprocessing, feature extraction and classification. Feature extraction is the most important part, and its quality directly affects the recognition results. The speech gender proposed by previous researchers Most of the features are based on the prosody features and sound quality features of speech, which are all artificially designed features. The robustness of the system is not high, and it is easily affected by the e...

AI Technical Summary

Problems solved by technology

[0003] The current speech signal recognition system usually consists of three parts, which are speech signal preprocessing, feature extraction and classification. Feature extraction is the most important part, and its quality directly affects the recognition results. The speech gender proposed by previous researchers Most of the features are based on the prosody features and sound quality features of speech, which are all artificially designed features. The robustness of the system is not high, and it is easily affected by the environment.

Images