The decline in both the price and size of sensors, and the fact that they can be embedded in various wearables (for example, clothes, jewelry, and furniture), has made them appealing for many purposes, including human activity recognition (HAR). Consequently, HAR has become a rather booming field; hence, the release of this book was timely.

The book is structured in two parts: Part 1, “Human Activity Recognition: Theory Fundamentals,” and Part 2, “HAR in an Android Smartphone: A Practical Guide.” This structure makes sense, but also signals limitations of the book. HAR systems are sensor-based; other techniques such as computer vision are completely ignored [1]. On the one hand, I like this, as it gives a genuine focus. On the other hand, it would have been good if some of the alternative and complementary techniques were mentioned. Applications are linked to smartphones solely, with one operating system (OS): Android. Good for those who like Android, but not so good for everyone else. Some of the chapters are based on articles published in several venues. This makes it less of a monograph than I had originally expected.

Part 1 includes six chapters. Following an introductory chapter, chapter 2 explains the fundamentals of HAR using sensors. The chapter starts with identifying the eight main design issues. Although obvious when reading them, a few of them are often ignored. Therefore, it is valuable that these fundamentals are described. Regrettably, in the signal processing pipeline, a core component is missing: feature or dimension reduction. Although often ignored, this remains important, in particular for real-time applications. More generally, I would argue that signal processing, which is crucial for HAR, is hardly touched. In the section on supervised learning, I missed random forests as a relatively new, powerful technique [2]. More advanced topics, such as handling imbalanced data, are ignored completely [3].

Chapter 3 simply provides a listing of HAR systems, which is nice to have but adds little to the book. In chapter 4, the authors describe the use of physiological signals for HAR purposes. This chapter describes a study instead of providing a solid overview of this area. Moreover, it does not touch on related fields that use physiological sensors, such as affective computing [4] and biometrics [5].

Real-time activity recognition is discussed in chapter 5. It describes a study, but does not redeem its promise. The real-time aspect is not treated convincingly. Issues such as computational complexity, parallelization, and big data analytics (for example, using Scala and Mahout) are not discussed. Chapter 6 explores multiple classifier systems (MCSs). This seems to be a pragmatic choice, with perhaps some work being available already. Finding such a topic in a book like this surprises me since so many more fundamental issues are left untouched.

Chapter 7 is the first of eight chapters in Part 2. In this chapter, the authors offer some conclusions for the first part of the book; these have limited value. Chapters 8, 9, 11, and 12 provide an introduction to Android. This is rather misplaced in a book such as this, mainly because it has nothing to do with HAR. Chapter 10 is on using smartphone sensors, which is essential for HAR. Regrettably, this chapter provides only an example application. Chapters 13 and 14 discuss feature extraction and machine learning (using the Waikato environment for knowledge analysis (WEKA) [6]) using a smartphone. This is interesting and of high practical value for HAR.

Taken together, this book’s topic is definitely hot. It has its charm, with a good introduction to the fundamentals of HAR. Some practical suggestions on feature extraction and machine learning using a smartphone are also of value, although these topics are of a more generic nature and not specifically tailored to HAR. This raises the following question: What does this book add when compared to a signal processing and machine learning introduction plus a survey article on HAR [7]? Little to nothing, I would pose.

More reviews about this item: Amazon