The formal idea of a home automation system (HAS) originated during the World's Fairs of the 1930s and has been a topic for many science fiction writers. For example, H.G. Wells wrote about automatic doors in The Sleeper Awakes published in 1899. Utopia, as theorized by science fiction writers and political ideologists alike, was associated with the idea of mitigated work through the use of machines and many automation methods aim to reduce the amount of work required to perform normal, often considered menial, tasks around the home. As case in point, wireless remotes could be used for vessels and vehicles to keep a person from having to physically be present to manipulate the device, an idea that Nikola Tesla patented in 1898.

Automation is already a common facet in a first-world home as seen in washing machines to clean clothes, dishwashers to wash dishes, water heaters to remove the need to manually heat water for bathing, and thermostats to automate the temperature. These technologies are only a fraction of the devices that can be automated. One of the most famous homes, the home of Bill Gates, expounds on the possibilities that lie in home automation. Each person is pinned with an electronic tracking chip and as a person traverses the home lights turn on ahead of the chip and turn off behind it. The chip maintains data on everything a person does and makes adjustments as it learns that person's preferences, including musical tastes and television channels. When two people pinned with such chips enter a room the system attempts to compromise between the preferences of both.

A "smart house," or a home that is equipped with lighting, heating, and electronic devices that can be controlled remotely as coined by the American Association of Housebuilders in 1984, began as a wired electrical system intended to provide power to multiple outlets, intended for a television, lighting, and a doorbell. 'Remotely' used to refer to another room located in the same domicile, but now the term 'remotely' has extended to include any location where an Internet-connected tablet or cell phone can reach. The boom of home automation is closely linked to the ability to control it from a tablet or touch screen device. Smart homes now contain any of the following automations: kitchen appliance and lighting, lawn watering and care, pet feeding, security systems, camera systems, audio and entertainment systems, window coverings, and utility usage to include both water and electric.

A speech recognition system simplifies a user’s interaction with a home automation system (HAS). Speech recognition applications can be classified into three broad groups of isolated word recognition systems: each word is spoken with pauses before and afterwards, small vocabulary command and control applications, and large vocabulary continuous speech systems. A smart home system uses the mixture of the second and third cases. The main feature of a smart home is the ability to remotely command the functions of the home via its appliances. For example, controlling the amount of light in a room by a light switch. This is the command and control function. Most command and control applications have a small vocabulary size that reflects the operations required to control the equipment. For example, “television off” would turn the television off. However, more complex commands can be managed through a known set of alternatives, such as days of the week, percentages, or times of the day. As the number of alternative wordings increases, the task of listing all possible combinations and associating them with a given set of actions become unmanageable and so a grammar syntax is required that specifies, in a more abstract way, the words and phrases along with their permissible combinations. Minimizing the vocabulary used to control devices also allows the SCREAM system to minimize its size and cost.

The SCREAM system runs on a secure network, controls door locks, and provides a signal which manipulates power to electronic devices without using a core visual interface like a tablet. Vocal control is possible through a wired headset or wirelessly through an Android phone and a third party Bluetooth device provided it can connect to the audio port on the main unit. While speaking the SCREAM system can recognize speech commands, interpret them, and control the appropriate device. SCREAM is less expensive than other systems because there is no GUI, which removes the need for an expensive LCD display, and the number of electronic control units are scalable. The system runs on low power to ensure that although the system provides these features, it will not add significantly to a utility bill. Due to the popularity of home automation technologies it was necessary to keep security in mind to prevent unwanted entities from gaining access to the premises. Lastly, the system was easy enough to set up and use for the average person with minimal computer knowledge.

Speech Recognition Circuit (SRC)

The heart of the speech recognition system is the HM2007. It is a single chip CMOS voice recognition LSI circuit with on-chip analog-to-digital conversion (ADC) which is used for voice analysis and recognition. Using 64K of SRAM a maximum of 40 words, each up to 0.96 seconds in length, can be trained. Alternatively, the device can be configured to accept 20 words up to 1.92 seconds in length. Using an external microphone and number pad the circuit can store voice recordings which can later be recognized as matching words. The circuit has a response time of less than 300 ms and can operate with a single 5V power supply. The SRAM has a 3V battery back up used for memory retention should power ever be removed from the main circuit.

The HM2007 IC is conceptually designed to be able to run in two different modes: Manual and CPU. In manual mode, a 4x3 button keypad is used to input command numbers during the training sequence. The keypad uses a keyboard scanner which allows the required number of I/O pins to be reduced from 12 to 7. CPU mode interfaces those same I/O pins with a micro controller and allows commands to be sent and statuses to be read using on-chip input and output registers. The advantage of CPU mode is that software can handle inputting command numbers and free the user from having to use the keypad. However, there is a well documented flaw with the IC which prevents use of CPU mode. For this reason, manual mode was used and keypad entries are still handled by the user. To alleviate the burden of having to look up command codes, they have been provided to the user in the systems interactive training menu.

Microcontrollers

Originally, the CC3200 by Texas Instruments was intended to drive all of the modules in SCREAM, including the main module with the SRC. However, it was determined that the CC3200 could not accept more than one connection once established as an access point (AP). This was unacceptable for the purposes of SCREAM as the main module must accept connection and communicate with multiple modules. Upgrading the microcontroller in the main module was necessary both to handle an unexpected increase in the required number of I/O pins and to ensure an AP could be created to handle the network traffic to the outlying modules.

The ODROID-C1 is a Raspberry Pi derivative created by HardKernel. Unlike the Raspberry Pi, however, the ODROID- C1 provides more computational power and faster RAM for the same price. The standard USB ports available allowed the ODROID-C1 to be equipped for wireless communications via a Wi-Fi dongle and the ethernet connection allowed for easier setup and troubleshooting via Wireshark. The 19 available general purpose I/O (GPIO) pins were enough to drive the SRC, the LCD, and the rotary encoder. A 16 GB flash secure digital (SD) card provides space to save both the operating system and any additional content for the ODROID. SCREAM does not require a graphical user interface so instead of using Ubuntu 14.04 as the Linux platform, an ARM distribution of Arch Linux was used instead. Arch Linux is a minimal installation distribution meaning that only the services required for the system to be operational needed to be installed. This conserved space on the SD card that was later used as available space for scripts and configuration files.

As the CC3200 could still establish itself as Wi-Fi station using WPA encryption and operate as a HTTP server, it remained the driving microcontroller for the outlying modules. The CC3200 is a Wi-Fi certified single chip microcontroller unit (MCU) with built-in 802.11 b/g/n Wi-Fi. With an ARM Cortex-M4 core it provides up to WPA2 encryption and embedded proprietary TCP/IP and TLS/SSL stacks. Sockets are established using the Berkeley socket API. Software written for the CC3200 can be compiled and debugged in Code Composer Studio or Energia and can be flashed to the available 16 MB of space using TIs proprietary software, Uniflash.

Door Control Unit

A normally-open (NO) mode fail-secure electric deadbolt is the fundamental part of the door control module. The electric deadbolt consist of an electromagnet and an armature plate. The locking device is fail-secure, which means the device remains locked if power is lost. The armature plate is attached to the door and the electromagnet portion of the lock is attached to the door frame. The unit operates at 12 W of power (12 VDC, 1 A) and is powered by a lithium-ion rechargeable battery. The rechargeable battery supplies a nominal voltage of 12 VDC with a capacity of 3800 mAh. The CC3200 launchpad permits the wireless communication between the speech recognition circuit and the door module. The CC3200 behaves as a transceiver receiving an input signal when the user interfaces with the system and provides a corresponding output of 3V. To control the locking/unlocking functionality of the electric deadbolt, the TIP120 darlington transistor is incorporated with the CC3200. Depending if a high or low output signal is supplied from the CC3200, the transistor is either set to saturation state or cut-off state, respectively. The CC3200 manipulates these states which makes the transistor act as a current control switch, and thus sets the electric deadbolt to be locked or unlocked. The electric deadbolt works by magnetic induction. When the system is suddenly turned off it might create a high voltage spike signal that can cause damage to the circuitry of the system. The IN4005 rectifier diode is used to handle the current and protect the unit from a high output voltage peak coming from the solenoid of the bolt when system turns off.

Light Control Unit

The CC3200 in the relay control unit processes received signals and when necessary applies or removes voltage from GPIO pin 29. This I/O pin is also connected to an onboard LED so a change in status can be noticed even if no electrical device is connected. The CC3200 launchpad operates with two AA batteries, which allows the CC3200 to send a high level signal of 3V and 10.75 mA through pin 29. Voltage is applied/removed across pins 29 and 20 (GND pin) to send a high/low level voltage signal to the relay control unit’s (RCU) printed circuit board (PCB). The board houses a high power switching TMP relay SPDT type, a P2N2222A bipolar NPN switching transistor, a rectifier diode 1N4005, a 270 Ω resistor, a three pin female header power connector, and rechargeable 9 Volt Lithium Polymer Battery.

More information can be obtained from the downloads page in our IEEE conference paper or the senior design II paper.