Advanced Controls

You execute this option in two ways. It is part of the ACORNS tools menu. It is also part of the SoundEditor application. If you are in the SoundEditor, simply press the button. When you do this you will see the following dialog frame. When the controls frame is executing, You can change any of the fields that show. If you move the mouse over any of the text boxes, you will see the legal input values. If you change these fields, the ACORNS sound editor will use the modified values from then on. It will remember the changes when you next run the program.

If you are not familiar with digital signal processing, some of these fields might not make sense. Don't worry. The default values are normally fine. The sections below give a brief description of each field.

Audio Compression Extension ACORNS presently can decode nine audio input formats (mp2,mp3, au, aif, aifc, ogg, wav, spx, and gsm) and can encode in six of those. The default encoding is wav because it is recognized on all platforms. However it uses quite a lot of memoty. You might experiment with the other settings to determine which you prefer. For example spx is good for speech because the level of compression is approximately twenty five to one, with accurate playback.
Recording Rate and Device On Mac systems the recording rate must be 44100 samples per second so this is the default. Window systems can handle other rates. Normally, speech only requires 16000 samples per second for acceptible playback.

The recording device selection is for systems that have multiple devices to record. If ACORNS does not correctly pick the one desired, you can alter the one it uses.
Voice Activation Window Size and Threshold ACORNS has a voice activation feature that will prevent it from recording audio until the person with the microphone begins to speak. By default this option is disabled. If you enter a positive decibel threshold (normally between 30 and 40 for speech) you can activate this feature. The voice activation window size indicates the number of milliseconds at a time that ACORNS will check to see if there is any speech. A large number could make the voice activation feature less reliable; however, you might be able to be increase it somewhat.
Frame Rate This field controls the sample rate ACORNS uses internally for audio fields. The 16000 rate is sufficient for most playback and speech recognition applications. Higher sample rates will save in larger files and require more processing, and download times.
Bits per Sample Sound data normally stores using either eight bit or sixteen bit samples. Eight bit formats are smaller, but can have poorer sound quality.
Channels? A value greater than one indicates that sounds will record in stereo. Speech data normally does not require stereo. Stereo files are at least double in size than sound recorded with a single channel.
Big Endian? This field requires a yes/no answer. That is why it shows with a question mark. PC compatible computers create files in little endian format. Motorola computers use big endian format. Standard file formats normally support both modes.
Encoding This field determines how the Sound Editor encodes the sound file. The standard values are unsigned, signed, alaw, and ulaw. Be careful. If you pick a format that is invalid for the type of sound file, recordings won't save successfully.
Averaged? This field requires a yes/no answer, hence the question mark. It applies to sound files recorded with multiple channels (stereo). A yes means that the sound editor will average the amplitudes of the channels. A no means that the sound editor will use only the amplitudes from the first channel.
Signal to Trim from the Front Sometimes sound files have a spike of high amplitude frames in the front. This can cause the normalize option to appear to not work. This option automatically silences the front portion of the sound wave to avoid this problem.
Round to Next Zero Amplitude? This field requires a yes/no answer, hence the question mark. When applying the various editing options, it is commonly advisable to find the next zero crossing point in the sound signal. This way, unexpected clicks can be avoided. It is a good idea to keep this field set to yes.
Maximum Recording Length Long recordings use lots of storage and can be expensive to process. For these reasons the ACORNS Sound Editor puts an upper limit on recording size. The default limit is ten minute sound clips. You can override this default by typing a different number of minutes into this field. Be careful. A value is that is too high can cause the sound editor to crash on systems with limited resources.
Window type The Sound editor provides a choice of window algorithms for its signal processing algorithms. This option allows the user to choose among these algorithms.
Filter Size Long filters do a better job but require more processing. If you are unconcerned with the filter type, leave this field alone.
Filter Type The frequency domain and spectrogram displays use this field. Digital signal processing uses filters to eliminate unwanted data from the signal. This option will not affect the time domain display or any of the program's editing options.
Preemphasis Factor In speech recognition algorithms it is often difficult to recognize high frequency sounds. This preemphasis algorithm modifies the speech signal to emphasize the higher frequencies. The value in this field determines how much emphasis is to be done.
Window Size When a speech signal is chopped into sections, the rough edges cause problems for speech recognition. Applying a hamming window algorithm smooths out the edges. The size of windows need to be small enough to reflect the changes in speech, yet not too small. This field sets the window size in milliseconds.
Window Shift Speech recognition algorithms break up the speech into overlapping sections called windows. This field determines how much overlap is desired.
Linear Prediction Algorithm The Sound Editor uses linear prediction for filtering, for display, and for beta speech recognition algorithms. There are several ways to calculate linear prediction and this gives the user the ability to choose among these approaches.
Linear Prediction Coefficients The number of linear prediction coefficients impacts the accuracy of the algorithm. This is a user selectable value.
Spectrogram The Sound Editor supports both wide and narrow band spectrogram output of speech signals. The approach used is determined by this parameter.
Color Palette The Sound Editor audio display can be either in color or use a grey scale. This parameter allows the user to choose the preferred approach.
Rate Change Algorithm There are various ways to alter the rate of speech to be faster or slower, and still maintain accurate pitch. Presently, the Sound Editor only supports the pitch synchronous overlap and add (PSOLA) algorithm.
Pitch Detection Algorithm Determining accurate pitch is important to many digitial speech processing problems. This parameter allows the user to choose among supported algorithms.
Rate change percentage per click When the user clicks on the icon to speed up or slow down speech, the rate adjusts by the percentage designated by this parameter.
Number of Mel Frequencies Human hearing is more sensitive to lower frequencies than to higher ones. The frequencies in a speech signal are converted into Mel Frequencies so that the speech recognizer can see a signal that closely mimics how humans hear. This is done by an algorithm that uses a group of filters to manipulate the signal. This parameter controls the number of these filters that the speech recognizer uses.
Mel Filter Bank Type and Gaussian filter spread These parameters determine how a bank of Mel filters overlap. It is an important parameter that affects the quality of automatic speech recognition.
Number of Cepstral features Cepstral feature are a series of numbers extracted from each portion of a speech signal (called a window). This field determines the number of values that are present in each feature vector.
Feature Vector Mask The speech display shows the feature vector values displayed over time. This field is useful to be able to control which feature vector values display. One needs to understand binary to correctly apply this field. One bit pertains to each feature vector value. A one means that the corresponding value gets displayed.
Minimum Bandpass Frequency The above filters eliminate frequencies from the sound signal that are below a designated value. If you are not concerned with the filter type, leave this field alone.
Minimum Bandpass Frequency This parameter sets the lowest frequency that the Mel filters will consider. Since normal speech soundwaves have frequencies below 8192 cycles per second, high frequencies can normally be ignored.
Dynamic Time Warp parameters These parameters determine how ACORNS will compare two audio signals to determine whether they match. We advise that you leave these values alone.