Adding audio capabilities to TIAGo service robot

Audio Database for TIAGo Service Robot

“…We have focused our attention on the audio capabilities of the TIAGo and we have obtained good outcomes in [2,3,4,5,6]. In [2] three types of features have been used (Linear Predictive Coding -LPC, Linear Predictive Cepstral Coefficients -LPCC and Mel-Frequency Cepstral coefficients -MFCC) and eight classifiers were tested (Bayes Network -BN, Quadratic Discriminant Analysis -QDA, Support Vector Machines -SVM, Multilayer Perceptron -MP, k-Nearest Neighbor -kNN, kStar, Fuzzy Lattice Reasoning -FLR, and Random Forests -RF). Overall correct classification rates greater than 99% were attained using MFCC (from 20 to 38) features and SVM as a classifier.…”

“…Based on the results reported in our previously works, we have concluded that the TIAGo can be used as a service robot for assisting elderly or chronically ill people. That is why the aim of this paper is to improve the number of audio signals in our initial database from [2]. We have focused on improving the number of acoustic signals in the voice part of the database with signals corresponding to medicine names and also with frequently used words in the home environment.…”

“…The initial version of the audio database [2] consists in 1380 audio signals. We have added 1920 more audio signals, in order to improve the database.…”

“…The audio signals in the database were recorded with both TIAGo service robot and the simulated service robot as described in [2]. The acquisition block diagram used for the audio recordings using TIAGo is illustrated in Fig.…”

“…2. The recorded sounds were chosen to belong to the same scenarios as in [2]. Each data stream was recorded with 48 kHz, 16 bit-accuracy https://www.degruyter.com/view/j/cjece (such that to contain only one class of acoustic events), in order to be in accordance to the previous audio signals in the database.…”