Cough sound automatic identification algorithm, device, medium and equipment

Abstract

The invention provides a cough sound automatic identification algorithm and device, a medium and equipment. The algorithm comprises the following steps: obtaining a to-be-identified sample; performing endpoint detection on the sample, intercepting a fixed-length sequence with the duration time being t from a starting point, and setting the fixed-length sequence as a signal sequence and subjected to normalization processing; utilizing linear predictive coding to obtain a predictive signal sequence and a residual signal sequence; solving an MFCC parameter mfccr of the prediction signal sequence and short-time energy enr of the residual signal sequence, and constructing a feature vector; and inputting the feature vector into a linear SVM classifier, and judging whether the to-be-identified sample type is a cough sound sample or a non-cough sound sample according to the output of the linear SVM classifier. The algorithm overcomes the problem that the time domain characteristics of the excitation source are easily influenced by sound channel resonance, and has good recognition capability and high recognition accuracy.

Description

technical field [0001] The invention relates to the technical field of medical equipment and medical signal processing, and more specifically relates to an automatic cough sound recognition algorithm, device, medium and equipment. Background technique [0002] Coughing is a natural reflex action of the body and is a common symptom of respiratory diseases such as asthma, pneumonia, laryngitis and chronic obstructive pulmonary disease. In clinical practice, doctors often use information such as the frequency and intensity of a patient's cough as an important basis for diagnosis, but this information is often mainly derived from the patient's subjective description and has poor reliability. Different from the patient's subjective description, the recording and automatic recognition of cough sounds help doctors obtain more objective and accurate information about the patient's cough. [0003] One of the main difficulties in the automatic recognition of cough sounds is that ther...

AI Technical Summary

Problems solved by technology

Among them, the algorithm uses linear predictive coding (LPC) to realize the separation of the channel model and the excitation source, so as to overcome the difficulty in extracting the characteristics of the excitation source due to the influence of channel resonance; The temporal energy is used as the feature of the cough sound automatic recognition algorithm, which contains both the vocal tract model and the excitation source, so as to solve the problem that common algorithms ignore the characteristics of the excitation source; the algorithm uses support vector machines (SVM) as Classifier to overcome the problem of model training caused by insufficient samples and unbalanced samples; in addition, the algorithm only uses the first stage of cough sound to extract features, the model is simple and the amount of calculation is small, and it is easy to implement in wearable devices

Images