JPH04301696A - Audio signal processor - Google Patents

Abstract

PURPOSE:To detect the pitch of an audio signal with a simple device. CONSTITUTION:A comparison computing element 24 detects the conversion state of supplied amplitude data, and detects the pitch by the fact that data rises from now(the amplitude data is changed from a negative value to a positive value) and amplitude is sufficiently low. Therefore, such detection can be performed with a comparatively simple device. When fast listening is performed, the data is separated at every detected pitch, and processing is performed, thereby, it is possible to suppress the occurrence of an incongruous noise.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal processing apparatus for shortening or extending the reproduction time of an audio signal.

0002

2. Description of the Related Art Conventionally, various types of audio signal recording and reproducing devices are known and used in various fields. especially,
Semiconductor memory requires no mechanical drive means, is small,
Because it has the characteristics of being lightweight and low power consumption,
By utilizing this, the applications of recording and reproducing devices are becoming more and more widespread.

[0003] When playing back with a recording/playback device, there are cases where it is desired to listen to the recorded content in a short time or to listen to it over an extended period of time. For example, when playing recorded content on an answering machine, functions such as fast listening to hear the content quickly and slow listening to confirm the content of unclear audio are required. However, in such a case, if the audio signal is simply thinned out or added by interpolation, there is a problem that unpleasant noise will occur.

[0004] This is because even though normal sounds such as human voices can be heard as one sound, they do not continue as a constant amplitude sound for a long time, but rather have a predetermined period (this is called a pitch period). This is because the sound repeats its strength (called pitch), and if the sound is divided into sections at areas with large amplitudes, the locations of the divisions will become unnatural. For this reason, if the audio signal is subjected to segmentation processing for each pitch period, it is possible to minimize the adverse effects of the processing and perform listening such as fast listening or slow listening. That is, by shortening the thinned-out audio signal for each pitch period, the negative effects of thinning can be minimized, and the content becomes easier to hear during fast listening. Furthermore, if the audio signal is expanded by interpolating the audio signal for each pitch period, the content becomes clear during slow listening.

[0005] Conventionally, therefore, audio signals have been processed in accordance with the pitch period using the following method.

(a) Determine the autocorrelation of the audio signal and determine the repetition period (pitch period) from this.

(a) A period that is considered to be a general pitch period (about 10 to 30 msec) is estimated as the pitch period, and the pitch period is fixed to this.

[0008]

[Problem to be solved by the invention] However, the above (
In method a), the autocorrelation must be determined each time the process is performed, and the calculations involved are enormous, resulting in an excessive burden on the device in terms of both hardware and software.

On the other hand, although the method (a) in which the pitch period is fixed is very easy to process, there are problems in that there are many errors, the sound is cut off at high amplitude points, and unpleasant noise is generated. Ta.

0010

Means for Solving the Problems An audio signal processing device according to the present invention is characterized by the following.

[0011]

[Operation] The audio signal processing device according to the present invention has the above-mentioned configuration, and detects the pitch period through a simple process of comparing the change state of the audio signal. Therefore, processing is simple and can be performed at high speed with a small device. Furthermore, since the pitch period is actually detected, the occurrence of unpleasant noise can be suppressed to a minimum.

0012

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the audio signal processing apparatus according to the present invention adapted to fast listening will be described below with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of the device, and an audio memory 10 stores encoded digitized audio signals. That is, the audio signal input from a microphone or the like is digitized, encoded based on a predetermined protocol, and stored.

A memory controller 12 is connected to this audio memory 10, and this memory controller 12 has the following functions:
Controls recording and reading from the audio memory 10. The memory controller 12 also includes a sequencer 14.
is connected to control read timing, read start address, etc.

On the other hand, the data output line of the memory controller 12 is connected to a 1-word register 20 and a k operation section 22.
are connected, and the outputs of the 1-word register 20 and the k calculation unit 22 are input to the comparison calculation unit 24. Then, the comparison calculation unit 24 calculates the pitch of the audio signal from the input data, and supplies the calculation result to the sequencer 14. Therefore, the sequencer 14 controls reading of data from the audio memory 10 by the memory controller 12 according to the calculation result, and inputs a decoding control signal to the decoding processing section 30. On the other hand, the decoding processing unit 30
The data from the memory controller 12 is input to the , and the decoding processing unit 30 performs the following according to the decoding control signal.
Performs data decoding processing.

The data decoded by the decoding processing section 30 is outputted via the FIFO memory 32, converted to an analog signal, and reproduced as a normal audio signal. Note that the read frequency fs from the FIFO memory 32 is set to be the same as the conversion frequency of a DA converter that converts digital data output from the FIFO memory 32 into analog data.

Next, the processing operation of this apparatus will be explained based on the flowchart of FIG. First, the sequencer 14 turns on the decoding operation in the decoding processing unit 30 using a decoding control signal (S1), and sets a variable c1 representing the number of processed data (number of words) to 0 (S1).
2). In this example, one word is encoded data representing one amplitude. Then, the memory controller 12 reads m words of data from the audio memory 10 and supplies it to the decoding processing section 30. Here, S1
Since the decoding process is ON in , the read m word data is decoded (S3) and the FIFO
Since the signals are outputted via the memory 32, they are sequentially reproduced as sounds.

On the other hand, the next data to be output from the memory controller 12 is sent to the word register 20 one word at a time.
This is read out to the comparator 24 (S
4), where the pitch is detected. Here, the number of words m
is a sufficiently large number compared to the normal pitch period. Therefore, if pitch is not detected even after processing m words, 1 is added to variable c1 in order to forcibly consider the m word breaks as pitches (S5),
It is determined whether c1>m (S6). Then, only when c1<m, a conditional determination is made as to whether or not it is a pitch (S7).

[0018] In S7, if it is determined that the pitch is reached, or if it is determined in S6 that c1 has reached m,
The sequencer 14 stops decoding in the decoding processing section 30 in response to the decoding control signal (S8). And c1
is reset to 0 (S9), and the next m words are read out with decoding stopped (S10). Therefore, no audio is reproduced for m words after pitch detection.

[0019] Then, m for which this decoding was not performed
It is determined whether the data following the word satisfies the conditions (S11 to S14). Then, it is determined whether the process is finished (S14), and if it is not finished, the process returns to S1 and this process is repeated until the process is finished. When such processing is performed, as shown in FIG. 3, first m words are reproduced, and then reproduction is performed until the condition is satisfied. If the condition is satisfied, the data for the next m words are unconditionally discarded and no reproduction is performed, and after the elapse of m words, a point in time at which the next condition is satisfied is searched for. Then, playback starts from the point when the conditions are satisfied. In addition, in the processing of m words, if the condition is not satisfied, m
Switch between playback and non-playback each time a word passes. Ensure that a minimum amount of regeneration occurs. In addition, if the above m word is set to about 5 seconds, the pitch period is usually 10 to 30.
Since the time is about msec, the data is reproduced approximately every 5 seconds, and fast listening can be performed by intermittent reproduction. Since the delimiter between playback and non-playback is the detected pitch, it is possible to suppress the generation of unpleasant noise here.

Here, the condition determination in S7 and S14 is performed based on whether the following three conditions are satisfied.

(a) a<b (b) a*b<0 (c) |a|+|b|<k/i For example, as shown in FIG. It is something that is repeated. The place where this amplitude becomes small is the break (pitch) of the pitch period. Therefore, when AD converting such an audio signal, the amplitude value at that time is converted into digital data at a predetermined sampling period. Then, this value is stored in the voice memory 10, but if this value is a value for the amplitude centered on 0, n and n output from the voice memory 10 are
The data at the +1 sampling point is an amplitude value having positive and negative signs like a and b shown in the figure. Therefore,
By taking points that satisfy the above-mentioned (a) and (b) for this value, it is possible to extract points where the amplitude value changes from negative to positive. Condition (c) indicates that the amplitude value is less than or equal to a predetermined threshold value, and the pitch can be detected by extracting a predetermined low amplitude point.

[0022] Also, this threshold value is determined as follows for the maximum amplitude k:
It is determined to be a predetermined ratio (for example, 1/16, i=16). Therefore, the determination of this maximum amplitude k will be explained based on FIG. 4. First, at the beginning, the maximum positive value that the audio data can take (for example, if the amplitude is expressed as 8-bit data, the maximum value that can be expressed with 8 bits) is set to k as the initial value. (S10
1). Then, a variable c2 for determining the timing of k update
and set the variable M, in which the maximum value is temporarily input, to 0 (S
102) Do.

Then, the data regarding the amplitude is read from the 1-word register 20 and inputted to the variable D (S1
04). Next, it is determined whether the read data D is less than or equal to the previous maximum value M (S105), and if it is larger, the current data D is input to the variable M to update the maximum value (S106). . Then, 1 is added to variable c2 (S
107), determine whether c2>L (S108), c2
If c2 is less than or equal to L, the process returns to S104, and the process of inputting the maximum value into M is repeated until c2 reaches L.

[0024] Here, this L is a predetermined value, and is set to be sufficiently larger than m indicating approximately one pitch described above (for example, 6 m). And when c2>L,
The value of M is assigned to k (S109), and the process returns to S202. Therefore, every L (for example, approximately 6 pitches), the maximum value at that time is input to k. Then, the value of this k is the following 6
The pitch will continue.

Note that, as described above, since k is initially set to the maximum value, the amplitude threshold k/i is a large value. Therefore, it is easy to find points that satisfy the conditions. From the next time onwards, suitable pitch detection can be performed using the actually detected k.

Further, in the above example, fast listening is applied to intermittent playback, but audio data can also be thinned out on a pitch-by-pitch basis. This will be explained based on the flowchart of FIG.

The decoding process in the decoding processing unit 30 is
N (S201). Then, after setting the variable c1 to 0, in this state, read out one word at a time (S203)
, it is determined whether the conditions are satisfied (S206). here,
Before this condition determination, add 1 to c1 (S204),
Determine whether this c1 has not reached a predetermined constant n (
S205). This constant n is a number corresponding to one normal pitch, and is set to correspond to about 10 to 30 msec. Note that this n may be set to a value approximately twice the previous c1.

If the conditions are satisfied and the pitch is detected, decoding is turned off (S207), c1 is set to 0 (S208), and the pitch is detected in the same way (S208).
S209-S212). Then, when the pitch is detected, the process returns to S201 and decoding is started. by this,
Playback can be alternately switched on and off each time a pitch is detected, and fast listening that shortens the audio can be performed.

[0029] Furthermore, by combining such data pitch-by-pitch reduction with the above-mentioned data thinning at predetermined time intervals, it is possible to perform fast listening in a further shortened time.

Furthermore, in the above example, only fast listening was explained, but if data is thinned out (decoding OFF) by interpolation from the previous and subsequent data, data can be inserted and the playback time can be extended. . Since the data is inserted through interpolation, the transition of the voice becomes smoother, making it possible to reproduce unclear words clearly.

[0031]

[Effects of the Invention] As explained above, according to the audio signal processing device according to the present invention, with a relatively simple configuration,
Pitch can be detected. Therefore, processing such as fast listening can be effectively performed using the detected pitch.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the overall configuration of an embodiment.

FIG. 2 is a flowchart showing the operation of the embodiment.

FIG. 3 is an explanatory diagram showing the state of data thinning.

FIG. 4 is an explanatory diagram illustrating setting of conditions.

FIG. 5 is a flowchart showing the operation of k calculation.

FIG. 6 is a flowchart showing the thinning operation for each pitch.

[Explanation of symbols]

10 Voice memory 24 Comparison calculator 30 Decryption processing unit

Claims (1)

[Claims] 1. An audio signal processing device that processes an audio signal by dividing it into predetermined units, the device comprising: means for detecting the amplitude of the audio signal; and means for detecting a state of change in the amplitude of the audio signal. An audio signal processing device characterized in that the audio signal is separated by a location where the amplitude of the audio signal is less than a predetermined value and the sign of the amplitude is reversed.