Classification of non-speech acoustic signals using structure models

In this paper we propose a universal strategy for the automatic interpretation of sensor signals. We focus on acoustic signals. However, any time series may be used. We assume that changes in an object's state cause a typical and reproducible change in the characteristics of the acquired sensor signal. In such cases we can train pattern recognizers basing on Hidden-Markov-Models or support vector machines with data recordings of different object states and use these classifiers to assess the state of identical or similar objects. Our approach assumes that the sensor signals consist of elementary signal events and some kind of syntax defining their temporal relation (much like a musical score defines the temporal relation between notes). It is capable of automatically determining both, the elementary events and their syntax from the training data. We present experimental results from seven different applications from the fields of non-destructive testing, bio and music signal processing

show abstract

“…This problem can only be solved by constantly adapting the Hidden Markov Models while measuring. For details see [7].…”

Section: Quality Assessment Of Sintered Gear Wheelsmentioning

confidence: 98%

Pattern recognition for sensor signals

Wolff

Tschöpe²

2009

2009 IEEE Sensors

View full text Add to dashboard Cite

show abstract

“…Having models to choose from, they decide for the class whose conditional probability is maximal (7) As is not known directly, we used Bayes' theorem in (7) to reduce it to the determinable conditional probability that the vector is observed given the class and the prior probability of the class itself. 7 As the probability of the feature vector does not depend on the class, it cannot change the decision and may be disregarded. The remaining term is called likelihood function.…”

Section: B Gaussian Mixture Model Classifiersmentioning

confidence: 99%

“…We have successfully applied this method to: 1) quality assessment [7]; 2) structural health monitoring of aircraft parts [8] and train wheels [9]; 3) life cycle analysis [10]; 4) identification of individual musical instruments [11], [12]; 5) biological signal processing (auscultatory blood pressure measurement [13]); and 6) one of its original applications: automatic speech recognition (e.g., [14]- [17]). …”

Section: Introductionmentioning

confidence: 99%

Statistical Classifiers for Structural Health Monitoring

Tschöpe

Wolff

2009

IEEE Sensors J.

View full text Add to dashboard Cite

Multisensor problems are important tasks in the field of structural health monitoring. By means of signals originating from different sensors, we have to make a decision about the test object. We describe a universal, largely problem independent method, which applies statistical classifiers in order to identify objects or assess their state. This work presents the results of a series of studies, which systematically investigated such approaches for a great variety of technical and biological signals. We give an overview on the theoretical background and describe two selected application examples.

show abstract

“…It ensures desired classification task with high classification accuracy. Figure 5 illustrates the process of extracting Mel-Frequency Cepstral Coefficient(MFCC) [4] features from the input audio.…”

Section: Modelingmentioning

confidence: 99%

“…K-means clustering is a method of cluster analysis which aims to partition n observation feature vectors into k clusters under the mean sequence distortion criterion. In this work, VQ models with codebook sizes K ranging 4 to 256 (4,8,16,32,64,128,256) were built. As shown in Figure 6, cluster training exemplars y in S i into K clusters with cluster centroids C i = c i j , j = 1, ..., K. In testing phase, MFCC vectors are extracted from the test audio signals.…”

Section: Figure 5 Mfcc Feature Extractionmentioning

confidence: 99%

Non-destructive acoustic based moisture measurement system for cashew nuts

Agrawal¹,

Nagaraj²,

Karthik³

2011

Etfa2011

View full text Add to dashboard Cite

A novel acoustic based moisture meter 1 has been designed with the specific purpose of measuring moisture in raw cashew nuts. The meter employs pattern matching techniques to determine the moisture content. The meter applies non-invasive, non-destructive methods and the techniques are easily scalable to perform online measurements suitable in an industrial set up. For test trials, a set 15-20 cashew nuts of varying size and weight were considered for our system training and testing. The feature vectors for our training algorithm were based on the acoustic signature of shell to shell impact of cashew nuts. Algorithmic implementation and the electronic and mechanical aspects of the system prototype are discussed. By this method, we were able to distinguish between the dry and wet cashew nuts with high correlation accuracy.

show abstract

Classification of non-speech acoustic signals using structure models

Cited by 9 publications

References 6 publications

Pattern recognition for sensor signals

Pattern recognition for sensor signals

Statistical Classifiers for Structural Health Monitoring

Non-destructive acoustic based moisture measurement system for cashew nuts

Contact Info

Product

Resources

About