Signal Processing - March 2017 - 71

with some high-end microprocessors having dedicated builtin DSP functions, which significantly speed up signal processing and increase the computational efficiency. The result
is that not only are finite impulse response (FIR) and infinite
impulse response (IIR) filters quite feasible on modern platforms,
but considerably more complex algorithms, such as fast Fourier
transform (FFT), discrete Fourier transform (DFT), neural networks, and even modern machine-learning algorithms, have
become possible.
In the automotive industry, the development of engine control and diagnostics is outpacing the development of the engine
"hardware" itself. Sophisticated algorithms based on modern
control theory and modern signal processing are being integrated into engine control and OBD because simple and intuitive
algorithms no longer satisfy stringent fuel economy and emission
requirements. The deployment of these advanced algorithms is
facilitated by many new sensors being added to engine systems.
The end result is that an abundance of raw signal data is available
for processing, and DSP is essential for processing and cleansing this data, performing signal detection and important parameter estimation needed for vehicular processes. In particular,
the automobile is a rotational machine and, as such, produces
signals with periodic characteristics, and DSP techniques, such
as frequency domain filtering, are both effective and essential to

Analysis Domain

Crankshaft
Angular Position

Time-Domain

processing the data so that engine control and diagnostics challenges can be addressed in a robust manner.
This article provides a survey of DSP applications in engine
control and OBD with a focus on practical automotive applications. The application examples we explore will mainly include:
■ engine misfire detection and related resonance noise removal
and rough road detection
■ individual cylinder fuel-air ratio imbalance detection for control and diagnostics
■ engine knock detection.
Our discussion will cover many DSP methods used in the
automotive industry, ranging from traditional to advanced signal processing techniques. The overview will focus primarily
on the most common methods currently being applied to solve
real problems faced by automotive manufactures. The discussion will conclude with a summary of DSP applications in the
automotive industry and highlight practical implementation
matters. Figure 1 provides an overall taxonomy of key application categories and techniques reviewed in this article.

OBD
OBD has become increasingly essential in modern automobiles. Today, all light-duty vehicles are required to be
equipped with OBD systems by the California Air Resources

Oxygen Sensor

Knock Sensor

Crank Angle
Domain

Frequency
Domain

Filtering-Based
Methods

Observers

Statistical
Techniques

Model-Based
Methods

Neural
Networks

Cylinder
Pressure Sensing

Ion
Sensing

Techniques

Figure 1. The taxonomy of DSP applications in engine controls and diagnostics.
IEEE SIgnal ProcESSIng MagazInE

March 2017

Table of Contents for the Digital Edition of Signal Processing - March 2017