Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

/pdf/beck-depression-inventory-iinituitenoyi-kao-cha-3143h0pleh.pdf

Beck Depression Inventory-IIについての一考察

Investigation of different speech types and emotions for detecting depression using different classifiers

Electroencephalography (EEG)–based studies focus on depression recognition using data mining methods, while those on mild depression are yet in infancy, especially in effective monitoring and quantitative measure aspects. Aiming at mild depression recognition, this study proposed a computer-aided detection (CAD) system using convolutional neural network (ConvNet). However, the architecture of ConvNet derived by trial and error and the CAD system used in clinical practice should be built on the basis of the local database; we therefore applied transfer learning when constructing ConvNet architecture. We also focused on the role of different aspects of EEG, i.e., spectral, spatial, and temporal information, in the recognition of mild depression and found that the spectral information of EEG played a major role and the temporal information of EEG provided a statistically significant improvement to accuracy. The proposed system provided the accuracy of 85.62% for recognition of mild depression and normal controls with 24-fold cross-validation (the training and test sets are divided based on the subjects). Thus, the system can be clinically used for the objective, accurate, and rapid diagnosis of mild depression.

EEG-based mild depression recognition using convolutional neural network.

Clinical depression or Major Depressive Disorder (MDD) is a common and serious medical illness. In this paper, a deep recurrent neural network-based framework is presented to detect depression and to predict its severity level from speech. Low-level and high-level audio features are extracted from audio recordings to predict the 24 scores of the Patient Health Questionnaire and the binary class of depression diagnosis. To overcome the problem of the small size of Speech Depression Recognition (SDR) datasets, expanding training labels and transferred features are considered. The proposed approach outperforms the state-of-art approaches on the DAIC-WOZ database with an overall accuracy of 76.27% and a root mean square error of 0.4 in assessing depression, while a root mean square error of 0.168 is achieved in predicting the depression severity levels. The proposed framework has several advantages (fastness, non-invasiveness, and non-intrusion), which makes it convenient for real-time applications. The performances of the proposed approach are evaluated under a multi-modal and a multi-features experiments. MFCC based high-level features hold relevant information related to depression. Yet, adding visual action units and different other acoustic features further boosts the classification results by 20% and 10% to reach an accuracy of 95.6% and 86%, respectively. Considering visual-facial modality needs to be carefully studied as it sparks patient privacy concerns while adding more acoustic features increases the computation time.

MFCC-based Recurrent Neural Network for Automatic Clinical Depression Recognition and Assessment from Speech

Depression is a common mental disorder and one of the main causes of disability worldwide. Lacking objective depressive disorder assessment methods is the key reason that many depressive patients can't be treated properly. Developments in affective sensing technology with a focus on acoustic features will potentially bring a change due to depressed patient's slow, hesitating, monotonous voice as remarkable characteristics. Our motivation is to find out a speech feature set to detect, evaluate and even predict depression. For these goals, we investigate a large sample of 300 subjects (100 depressed patients, 100 healthy controls and 100 high-risk people) through comparative analysis and follow-up study. For examining the correlation between depression and speech, we extract features as many as possible according to previous research to create a large voice feature set. Then we employ some feature selection methods to eliminate irrelevant, redundant and noisy features to form a compact subset. To measure effectiveness of this new subset, we test it on our dataset with 300 subjects using several common classifiers and 10-fold cross-validation. Since we are collecting data currently, we have no result to report yet.

Detection of depression in speech

Depression is a frequent affective disorder, leading to a high impact on patients, their families and society. Depression diagnosis is limited by assessment methods that rely on patient-reported or clinician judgments of symptom severity. Recently, many researches showed that voice is an objective indicator for depressive diagnosis. In this paper, we investigate a sample of 111 subjects (38 healthy controls, 36 mild depressed patients and 37 severe depressed patients) through comparative analysis to explore the correlation between acoustic features and depression severity. We extract features as many as possible according to previous researches to create a large voice feature set. Then we employ some feature selection methods to form compact subsets on different tasks. Finally, we evaluate depressive disorder severity by these acoustic feature subsets. Results show that interview is a better choice than reading and picture description for depression assessment. Meanwhile, speech signal correlate to depression severity in a medium-level with statistically significant (p < 0.01).

Evaluation of Depression Severity in Speech

Various problems caused by stress seriously affect individuals' physical and mental well-being and have been receiving an increasing attention in modern lives. Since traditional stress assessment methods are lack of objectivity, affective sensing technologies have been studied for years. As detecting stress in speech has the advantages of non-invasive, portable, fast, and less expensive, many explorations were conducted to build stress assessment models. To find out a proper acoustic feature subset for a specific reading pattern, we performed the experiments with 30 subjects by three reading patterns: vowel, figure and sentence. We utilized feature selection and classification techniques to automatically select acoustic features and evaluate performances. Results showed that there are interactions between reading patterns and stress levels on speech features. Although Stress levels can be distinguished in any pattern of them (vowel, figure, sentence), sentence is a better choice with the best classification accuracy 88.15%. Furthermore, Line Spectral Pairs (LSP) features are indispensable for vowel, Mel-Frequency Cepstral Coefficient (MFCC) features are more effective for figure and the combination of prosodic, LSP and MFCC features is more suitable for sentence.

Assessing stress levels via speech using three reading patterns

Many studies were proposed to discuss acoustic correlates of stress in recent years. Considering some inconsistent experiment results, we supposed that stress should be categorized into long-term and short-term stress in this topic, and the trend of short-term stress induced by workload may be affected by long-term stress. This study contains three parts: first, we proposed an acoustic feature set chosen by feature selection, which can be considered as a measurement of the level of long-term stress; second, we showed that this set is immune to short-term stress in stress classification tests; finally, we observed some particular voice features mentioned in previous researches in our experiment and the results may imply that short-term stress trend is in connection with the level of long-term stress. In short, long-term and shot-term stress should be discussed separately in future researches for clear and explicit conclusions.

Tianyang Wang

Papers

Investigation of different speech types and emotions for detecting depression using different classifiers

Detection of depression in speech

Evaluation of Depression Severity in Speech

Assessing stress levels via speech using three reading patterns

Feature selection and classification of speech under long-term stress