Notes On The Use Of "Open Air" Microphones with Voice Recognition Software.
by Peter Monahan ©1999-2002 iAutomate.com

Introduction

 

Since the introduction of Voice Recognition software, many adopters of home automation technology have benefited from the ease of use, and seamless integration that the products offer in controlling a variety automation interfaces. While communication with these products via desktop microphone, telephones in the home, and remotely by telephone is simple,  convenient, and effective, many customers desire the use of "Open Air" Microphones to talk to their home from room to room, without the need to pick up a telephone, or use a single desktop microphone.
.

This approach  can be successful, but proper installation and balance are critical to the success of the "Open Air" concept. The issue is not the effectiveness of the Voice Recognition software and imbedded speech engine, and whether the product will support such a design, but more a function of the microphone (and associated equipment) installation, and whether it can deliver the proper sound levels to the Software at a consistent range, volume, and clarity to offer the speech engine the intelligible dialog that it requires to "recognize" commands.

 

In considering the use of "open air" microphones an attempt was made to balance Cost Vs. Functionality.  A range of products were available that could enhance the overall sound quality that is delivered to the speech engine, but the costs were prohibitive for the average user. As demand for the concept grows (and we are certain that it will), we believe that more designs will evolve specific to this concept,  at a more cost effective price. In the absence of a product specific to this use, we have taken the initiative to work with a manufacturer to introduce the SCM-410 Automatic Mixer to the Home Automation Industry.

The challenges have been great, but we intend on delivering a product to market that is successful "out of the box."

In the meantime, let us share some of our experiences and understanding with you, as you prepare for this integration.

Operational Expectations

Our biggest concern with the technology in general at this point is managing the expectations of the end user. For some reason, you can easily explain  that if person wishes to record a song, perform on stage, or talk on the radio, that they must speak several inches from the microphone to achieve proper sound pressure levels without reflection or loss. Even speakers at a podium are never more than arms length from a microphone.
.

Lead singers never face away from the microphone when they perform, nor do they perform from even five feet away.

Home automation enthusiasts, however, expect that you can take the same technology, and force it to work from twenty feet away with a speech recognition product while the stereo is operating at a moderate level. Quite a challenge......You would be shocked how many people ask us if the microphone will work from ten feet away in a crowded room with music playing. If you can manage your expectations, then the products *will* perform.

Our Golden Rule: "If you can't touch the microphone, you can't speak to it"

MUST READ TOPICS:

Microphones On The Ceiling- Shure Strongly Advises "NO!

Critical Distance and Microphone Placement

The Myth of Microphone Reach

Predicting Speech to Background Noise Ratio at the Microphone

The Room

 

Before the sound gets to the microphone, it must travel through the enviroment. The enviroment is the single biggest factor in determining the performance of a microphone.

Every room of the home is different in it's acoustical characteristics. Walls, floors, ceilings, furniture and even window treatments affect the properties of the room.

The best results are achieved in an acoustically "flat" room. One without hard reflective surfaces, such as hardwood floors or ceramic tile.

We are not stating that one cannot achieve acceptable performance in an acoustically "bright" room, just that you need to lower your expectations for performance a bit.  Kitchens can be a tough room.

Placement of the Microphone(s)  "Flat" Rooms

Typically in a room that has flat acoustical properties, one microphone is sufficient. Care must be taken to direct this microphone away from sources of noise such as Televisions, Stereo Speakers, etc.  In this sense, a directional microphone is preferred. Results will always improve by moving closer to the microphone while issuing commands.

In a larger room, better results can be achieved by installing two microphones, but you must follow the 3 to 1 Rule. The 3:1 rule says to separate the mics by a distance of at least 3 times that of the mic’s distance from the sound source (The Person Speaking).   In this case, the microphone gain is adjusted to a lower level, and the gate (we'll talk about gates in a minute) should be adjusted to a higher level of rejection.

Placement of the Microphone(s)  "Bright"   Room

Kitchens, foyers and other rooms with hard surfaces demonstrate an   acoustically "bright" or reverberant room. The key to performance in this type of room is to provide higher gain before feedback or more "direct" sound, and limit ambient sound pickup.

This can be accomplished by using a hemispherical microphone, like the PZM-11, located closer to the person speaking. In addition, the microphone should be installed in a location that avoids ambient noise, including sources of noise such as refrigerators, dishwashers, sinks, etc.

Because a unidirectional microphone is preferred, the microphone should be installed and directed to be perpendicular to any sources of ambient noise, yet close to the source, or person speaking. The gain needs to be reduced, and the person speaking must be closer to the microphone.

As with large rooms, two microphones may be a solution in a reverberant room as well, but the same 3 to 1 rule must be followed, where the mics must be separated by a distance of at least 3 times that of the mic’s distance from the sound source.

Mixers

A mixer capable of Phantom Power from 12-24 VDC is required for use with the PZM-11 microphone. The mixer must be able to provide 40ma of current to each microphone input. 

The Microphone input should be gated, with "Filibuster" capabilities, such as with the Shure SCM-410, which permit only one active microphone at a time.

Let the Mixer do the Work !!

Gates

Gating is most important when utilizing multiple microphones. Gating is the process of opening the audio channel when the sound pressure exceeds the preset threshold. Without gating, the audio level reaching the recognition engine is an unpredictable variable, and the results will suffer. By closing the audio channels for microphones that are not in use, a constant sound pressure level can be delivered to the recognition engine without excess ambient noise.

The Shure SCM-410 is an excellent choice for Automatic Mixing and gating. There is a multi-color LED which Indicates when a channel is activated and when it is clipping. This  makes this design suitable for an installer as it can be adjusted for each room.

The possibility exists in an active  home with a lot of ambient noise where multiple gates can be open at one time. This can cause problems for the recognition engine.

In this instance, and automatic mixer is preferred, that will permit only one gate to be open at any time. this is often referred to as a "Filibuster" mode.  A mixer capable of these features generally has a logic input that allows manual overrides that will open a specific microphone channel when activated.

Tone Controls

While the PZM-11 microphone is tuned to roll off low frequencies, and enhance the range of the human voice, it is at times desirable to compensate further due to the acoustics of a room. In a reverberant or acoustically "Bright" room, it may be desirable to Boost the Bass response and Lower the High Frequency response of the microphone.

 

While this is not possible with the PZM-11, good results can be achieved with the PCC-170 microphone.

The SCM410 Mixer has adjustable compensation circuits for high and low frequencies.

Computer Hardware

Our experiences are that Processor speed and RAM make a considerable difference under adverse acoustical conditions. Some Soundcards limit the master gain control in hardware. For this reason we recommend the following  hardware:

350 MHz Processor
128 MB RAM
Fast Hard Drive, 7000 RPM or better,  capacity of your choice.

SoundBlaster SB-16 WAV Effects Sound Card

 

Keyboard, Mouse, Monitor, Video Card, CD, and Floppy of your choice.