Method for evaluating possibility of dysphagia by analyzing acoustic signals, and server and non-transitory computer-readable recording medium performing same

An audio analysis method for providing information on dysphagia, the audio analysis method comprising:

• obtaining an audio signal by using an electronic device;
• and
• generating information for the dysphagia by analyzing the audio signal;
• wherein the analyzing the audio signal comprises:
• extracting one or more onset signals from the audio signal, wherein the one or more onset signals have a preset length;
• converting each of the one or more onset signals into a spectrogram image;
• determining, based on the one or more spectrogram images, one or more cough signals using a cough determination model, wherein the determined cough signals include at least one explosive cough signal including an explosive phase and at least one non-explosive cough signal not including an explosive phase;
• only for the one or more cough signals:
• determining whether the non-explosive cough signal exists using a phase classification model,
• wherein the phase classification model is trained based on a training data set, wherein the training data set comprises data corresponding to a first type cough signal comprising the explosive phase and data corresponding to a second type cough signal not comprising the explosive phase, and
• wherein the training data set have a characteristic that a frequency band of a signal corresponding to a preset time period from a start point of the first type cough signal in the first type cough signal is higher than a frequency band of a signal corresponding to the preset time period from a start point of the second type cough signal in the second type cough signal,
• wherein the generating information for the dysphagia comprises:
• generating the information relating a risk of dysphagia in response to the non-explosive cough signal is exist.