Extracting Behavior Identification Features for Monitoring and Managing Speech-Dependent Smart Mental Illness Healthcare Systems

Abstract

Speech is one of the major communication tools to share information among people. This exchange method has a complicated construction consisting of not the best imparting of voice but additionally consisting of the transmission of many-speaker unique information. The most important aim of this research is to extract individual features through the speech-dependent health monitoring and management system; through this system, the speech data can be collected from a remote location and can be accessed. The experimental analysis shows that the proposed model has a good efficiency. Consequently, in the last 5 years, many researchers from this domain come in front to explore various aspects of speech which includes speech analysis using mechanical signs, human system interaction, speaker, and speech identification. Speech is a biometric that combines physiological and behavioural characteristics. Especially beneficial for remote attack transactions over telecommunication networks, the medical information of each person is quite a challenge, e.g., like COVID-19 where the medical team has to identify each person in a particular region that how many people got affected by some disease and took a quick measure to get protected from such diseases and what are the safety measure required. Presently, this task is the most challenging one for researchers. Therefore, speech-based mechanisms might be useful for tracking his/her voice quality or throat getting affected. By collecting the database of people matched and comparing with his/her original database, it can be identified in such scenarios. This provides the better management system without touching and maintains a safe distance data that can be gathered and processed for further medical treatment. Many research studies have been done but speech-dependent approach is quite less and it requires more work to provide such a smart system in society, and it may be possible to reduce the chances to come into contact with viral effected people in the future and protect society for the same.

Figure 1

Training in a speaker recognition system.

Figure 2

Speaker recognition testing system.

Figure 3

Principle shape of a closed set recognition system.

Figure 4

Principle structure of an addresser recognition system.

Figure 5

Gaussian mixture model with its feature space and corresponding 2-dimension.

Figure 6

Model for monitoring and management (assembled with the proposed model, Figure 7).

Figure 7

Proposed model for verification and authentication.

Figure 8

Spectrum of spoken digits.

Figure 9

Sampling of spoken digits.

Figure 10

Spectrum selection for complete spoken digits.

Figure 11

Efficiency of “SHUNYA” and “EK.”

Figure 12

Variation in the frequency spectrum of spoken digit “SHUNYA.”

Figure 13

Variation in frequency spectrum of spoken digit “EK.”