Non-linear features and feature selection algorithms for speech based prediction of body mass index (BMI)/
The current research explores the topic on prediction of BMI status (normal, overweight and obese) using speech signal (/ah/ sound). In this research work, the speech samples were collected only from volunteers (age varies from Twenty years to Forty years) and wavelet packet based nonlinear entropy...
में बचाया:
| मुख्य लेखक: | |
|---|---|
| निगमित लेखक: | |
| स्वरूप: | थीसिस सॉफ्टवेयर ई-पुस्तक |
| भाषा: | English |
| प्रकाशित: |
Perlis, Malaysia
School of Mechatronic Engineering
2017
|
| टैग : |
टैग जोड़ें
कोई टैग नहीं, इस रिकॉर्ड को टैग करने वाले पहले व्यक्ति बनें!
|
| सारांश: | The current research explores the topic on prediction of BMI status (normal, overweight and obese) using speech signal (/ah/ sound). In this research work, the speech samples were collected only from volunteers (age varies from Twenty years to Forty years) and wavelet packet based nonlinear entropy features (totally 372 features) were extracted. To test the efficiency of the proposed feature set, the combination of standard features was used, and several experiments were conducted. The results indicated that the proposed features were more efficient in predicting BMI status. Two feature selection algorithms namely, the combination of sequential feature selection algorithms (CSFS) (Backward, Forward, Individual, and Plus-l-takeaway-r) and Hybrid PSOGSA optimization algorithm (HPSOGSA) with the combination of Particle Swarm Optimization (PSO) and Gravitational Search Algorithm (GSA) were proposed in this research work To reduce the feature size dimensionality, the cost of classification/learning algorithm and to identify the most useful feature subset. The limitation of proposed method is that a small database was used and the number of females is less. |
|---|---|
| भौतिक वर्णन: | 1 CD-ROM 12 cm |
| ग्रन्थसूची: | Includes bibliographical references. |