Resources | LpR Article | Applications | Technologies | Controls | Smart Lighting + IoT | Aug 31, 2020

Offline Lighting Voice Controller — An Alternative to Voice Assistants

The historical evolution of light control by users started with a mechanical switch followed by potentiometer and dimming, touch panels and touch screens, gesture and movement detection control, and ending with voice. Voice control are mostly online speaker independent like Alexa, Google Home and Siri. But there are privacy and security concerns. Filip Michalski, Electronic Engineer, and Cezary Skotak, CEO, from Cezos describe a speaker independent offline voice controls system, explain the technology and discuss the pros and cons.

Artificial light was always linked with the need to control it. For many years it was only a simple on-off control, which was done by a switch on the wall or on the power cord. In recent years, the needs arising from the change in lifestyles combined with the development of technology have enriched the spectrum of available options.

The development of technology in conjunction with lifestyle changes meant that lighting control now plays the role of the most popular and cheapest function of an intelligent home or building.

A Short History on Controls

The development of electronics meant that the simplest lighting control of the on-off type, used in cable circuits, was enriched by the ability to control the power of light bulbs, which allowed for the regulation of their luminous flux. Although the maximum brightness of the lamps remained unchanged, thanks to electronic dimmers, it could be reduced, i.e. dim the incandescent lighting below the maximum level. There were some problems with it, such as the frequent acoustic effects of dimmers in the form of an annoying buzzing and the cost of the equipment itself, which significantly limited the spread of this control method. Electronic dimmers - slide, rotary knob or touch - were installed instead of the usual switches and could work with all traditional bulbs. Unfortunately, in the case of halogen bulbs there were significant problems in the form of a decrease in their durability. Dimmers were also not compatible with energy-saving compact fluorescent lamps, which largely displaced traditional incandescent bulbs from home use. In the next phase, wall touch panels began to be created, and thanks to things like home automation systems, it was possible to control the entire building from one point. The alternative control methods were infrared controllers and radio controllers, which could be controlled by a remote control.

The mass introduction of LED light sources for home use has become the limit of the possibility of using dimmers in the old, classic form. The overwhelming majority of LED lamps with the shape and size of traditional incandescent bulbs are incompatible with such old dimmers, and the price of LED lamps specially adapted for this is high. In turn, LED modules and strips require a special controller, which increases installation costs.

Further lighting control options, primarily LED lighting, appeared along with the mass spread of mobile phones: especially smartphones. Wireless lighting control systems have developed that use the Bluetooth standard for communication. Applications installed on smartphones have become a perfect, intuitive interface. This eliminated the need to approach the control board, because everyone could use their own, portable telephone usually kept within reach, to send commands wirelessly.

Despite their functionality in the mentioned drivers, the problem of the requirement to have a telephone or remote control at all times remained. These modern systems, in which the interface is a smartphone, require the user to be able to configure it. Some people who use mobile phones on a daily basis only use them for conversations and this is where their skills end. For them, such a configuration, although usually simple, turns out to be too difficult.

That is why systems with alternative operating methods are constantly being developed, which are becoming more and more popular. One example of a modern solution is the gesture control system, i.e. hand movements - ProxiDimmer. The controller, through a proximity sensor, can recognize the on-off gesture when moving the hand and lighten-dim when holding the hand over the sensor, and optionally change the color of the light. It is a very functional and aesthetic solution because it does not require touching the switch or panel. It also solves the problem of positioning the button, because the sensor can be placed next to the LED light source or even between the diodes (ProxiDimmer OnBoard), which allows for uniform light.

Figure 2: Proxi Dimmer On Board, gesture sensor with a LED module integratedFigure 2: Proxi Dimmer On Board, gesture sensor with a LED module integrated

From the user's point of view, the most convenient way to control the light, e.g. when his hands are occupied or when the switch, remote control or telephone are out of reach of hands, is to use voice commands. This is the most intuitive and fast control. We simply say the command specifying what, where and how we want to change, and the system recognizes and executes it.

Voice Control – The Technical Basics

This technology is not as modern as it may seem. It has existed for a long time, but it has always been limited by hardware and software capabilities. Despite this, the producers saw its potential. One of the first examples of its widespread use was voice dialing on mobile phones in the days before the popularization of iOS and Android. Unfortunately, the limitations of this technology were clearly visible and therefore it remained in the sphere of technical curiosities.

Comparison of Different Approaches for VC

The simplest voice control system is the Speaker dependent one. We need to teach the driver specific commands issued by our own voice, and then it will be able to recognize and execute them. The speaker-dependent voice control system is the cheapest and simplest solution. A good example of the location of such a system is a desk or bedside lamp. In this case, we usually deal with only one person controlling this lamp. The advantage of such systems is low power consumption, which allows battery operation, and no language barriers - the device only recognizes the pronunciation of a given command. It is limited by the need to configure commands that may seem tedious, and the limited number of them to configure.

Figure 3: Offline Speaker dependent voice controllerFigure 3: Offline Speaker dependent voice controller

VOICEEMITY Dynamic White Speaker Dependent controller parameters:
•    Operating voltage: 11 - 27 V DC
•    Application: Constant Voltage LED stripes and modules
•    Output current: 9 A/CH
•    Detection distance: 1 m
•    Number of commands: 8

A serious acceleration of control development using voice commands were provided by voice assistant systems – Amazon Alexa, Google Assistant, Apple Siri and Microsoft Cortana, connected to a suitable, intelligent device containing a speaker and a microphone, which, however, require access to the Internet, i.e. it is an online system. These solutions allow for many different activities, e.g. checking the weather or traffic conditions, and of course for voice control of lighting. All you need to do is buy a compatible light bulb or a system for LED strips or modules compatible with one of the mentioned voice assistants and you can enjoy the perfect control lighting system using your voice. These systems allow for various activities, e.g. switching on and off the light, grouping lighting fittings, programming light scenes, automatically switching the light on and off at a programmed time, etc. Is this the perfect solution? Yes and no. These systems require a permanent connection to the Internet, because it is in the cloud that our voice is analyzed. The device only reacts to the key word, so-called WakeWord, e.g. Alexa, Hey Siri or Ok Google, and then our commands are sent to the appropriate servers. There may be problems, because not only do you need to have efficient Internet access, but other elements of the chain must also work fine. The key here may be failures on the part of the voice assistant service provider or failures of the home Wi-Fi router. Lack of connection to the network or failure of another element means the loss of control over the light. An additional problem is the requirement to configure the voice assistant and each device - a light source or lighting fixture, and as we have already mentioned, it is a real challenge for many people.

Figure 4: Offline Speaker independent voice controllerFigure 4: Offline Speaker independent voice controller

HOMEEMITY Dynamic White Wi-Fi Voice controller parameters:
•    Operating voltage: 11 - 27 V DC
•    Application: Constant Voltage LED stripes and modules
•    Output current: 9 A/CH
•    Detection distance: depends on the voice assistant
•    Wireless system: Wi-Fi 2.4 GHz

The most interesting solution is a voice control system independent of the speaker (Speaker independent) that doesn't need the Internet to work (OFFLINE system). The system is always designed and built into one specific device, i.e. lamp above a table. Advances in hardware (price and capabilities) and software (reducing hardware requirements for power and memory) allow for matching quality with voice assistants that require the Internet. Of course, such a device can't predict what the weather will be like tomorrow or what the traffic will be like, but it has several key advantages.

Key advantages of speaker independent offline systems:
•    It doesn't require Internet, which makes it independent of it.
•    Complete lack of configuration, plug & play system with enough capabilities for light control
•    You can talk straight towards the device - a lamp that has its own built-in microphone.

As the previous systems are closed source – i.e. we don't know what is under the "hood", the speaker independent system is designed in collaboration with our voice recognition specialists. In this system two parts may be distinguished: key word spotting and command recognition. To minimalize delay and preserve consistency, both of those both tasks are realized with the same acoustic model. For acoustic modeling LSTM (Long short-term memory) a recurrent neural network is used. Network is trained using CTC (Connectionist Temporal Classification) cost function. To satisfy device performance, inference of the neural network is highly optimized with quantization, taking into consideration ARM SIMD.  Key word spotting is realized with CTC scoring algorithm. Command recognition operates on a beam search algorithm with limited vocabulary. Generic approach for acoustic modeling does not restrain domain and allows for high flexibility in target implementation.

Figure 5: Offline Speaker independent voice controllerFigure 5: Offline Speaker independent voice controller

VOICEEMITY Dynamic White Speaker Independent controller parameters:
•    Operating voltage: 11 - 27 V DC
•    Application: Constant Voltage LED stripes and modules
•    Output current: 9 A/CH
•    Detection distance: up to 5 meters

Table 1: Differences between the presented systems

How to Choose the Appropriate System

If someone wants to make a complete home automation system, an online voice assistant system would be a perfect choice. Ability to group lights, create automatic tasks (routines) is a great addition. It is also capable of many different tasks beyond light control, and thanks to constant updates, it is getting smarter. The requirement of constant connection to the Internet can be solved by adding an external switch or control option (i.e. gesture control), which would work parallel to the voice control.

If someone is concerned about data privacy, they don't want to rely on the internet connection or the configuration could be troublesome for you or your costumers, an offline system is an alternative solution.

With a speaker dependent system, the configuration could be somewhat of an issue, but it is still easier than a voice assistant. It is designed for lamps closer to the speaker (the detection distance is the lowest of all described systems) and most suitable for personal lamps i.e. desk or battery operated lamps

While the system above is not suitable for every lamp, the speaker independent solution is a solution for every lamp. No configuration, detection distances are the same as an online system, and privacy is guaranteed by design. Also worth noting, each lamp has its own microphone, so they can be put wherever they are wanted – there is no need to place a voice assistant system nearby to get the voice control. This technology enables the buyer to easily create smart lamps that are usable in everyday use and are easy to configure.

Conclusions

While every system has its advantages and disadvantages, the user itself needs to decide on the solution that is the most suitable for them. While the online voice assistant systems are constantly improving and work is being done to allow the user to control the devices (i.e. light) when the Internet is down, a parallel solution (Gesture, Touch Button and so on) is practically a must-have to be suitable for everyday use. This is true for every voice control system. While an offline voice control system can never replace an online system because of the added capabilities, it can be an interesting alternative.

Cezary Skotak:
Cezary SkotakCezary Skotak is founder and CEO of a family business.
The company, founded in 1981, is a producer of plastic and metal parts for the manufacturing of lamps.
In 2005, in strong cooperation with Osram OS, a strong partner in the LLFY platfrom, activity was expanded to include the production of LED modules and electronics to control them.
The year 2016 was the beginning of work on a new generation of voice-controlled lighting.

page_peel