JPH11187117A - Recording and reproducing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a high-reliability importance degree judging material concerning a recording and reproducing device for first outputting one part of a message as a message importance degree judging member. SOLUTION: A high volume part calculating means 7 extracts the high volume part of a voice message recorded on a voice recording medium 3. A pause part calculating part 8 detects a pause part in the recorded voice message and extracts a part sandwiched between long pause parts in the voice message. A speaking speed calculating means 10 calculates syllabale speaking speed in the recorded voice message and extracts the part of low speaking speed in the voice message. Then, an extraction part determining means 11 determines the part to be outputted as an important part in the message out of the parts extracted by the high volume part calculating means 7, pause part calculating means 8 and speaking speed calculating means 10.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a recording / reproducing apparatus which can store a message from another person, such as a voice mail or an answering machine, and reproduce it when necessary. The present invention relates to a recording / reproducing apparatus capable of determining importance.

[0002]

2. Description of the Related Art A recording / reproducing device such as a voice mail or an answering machine records a message on a recording medium by automatically responding to access from another user when the user is absent. To listen to recorded messages directly or remotely. However, each time a message is inserted, the message is sequentially recorded on a recording medium. Therefore, when reproducing, the user needs to listen to all the messages in order. Therefore, when a large number of messages are recorded and the last message is very important, it takes a long time to listen to the message. Further, when listening to a message by operating from a remote place, it may be difficult to hear a large number of messages due to a high communication cost such as a telephone charge.

In order to solve these problems, it is conceivable to shorten the time required for reproducing a message. For example, if the message is played back at twice the speed, the playback time can be halved, so that the user can arrive at the important message earlier. However, increasing the reproduction speed makes it difficult to hear, so there is a limit to the increase in the reproduction speed, and this is not a fundamental measure.

As another method for arriving at an important message early, there has been proposed a method of first listening to a part of each message to determine the importance of the message and then listening to the whole message only when necessary. . For example,
JP-A-63-179652 and JP-A-3-1860
The answering machine disclosed in Japanese Patent Publication No. 46 is to separately record a message to which the other party answers and a message,
At the time of reproduction, it has a function of reproducing the part whose partner has answered the name first. The user can determine who the message is from first, and based on the determination, determine whether the message is important.

[0005]

According to such a method, even if the entire length of each message is long, the content of each message can be determined to a certain extent in a very short time by reproducing a certain part. . However, it is often difficult to judge only the name of the other party, and it is difficult to determine whether the message content is truly important only by the name of the other party.

Accordingly, an object of the present invention is to provide a recording / reproducing apparatus which enables a user to determine the importance of a message based on highly reliable importance determination material.

[0007]

SUMMARY OF THE INVENTION A recording / reproducing apparatus according to the present invention records an input voice message and outputs each voice message recorded and a part of the message in response to a user's request. High volume portion calculation means for extracting a high volume portion in the recorded voice message; and a pause portion calculation for detecting a pause portion in the recorded voice message and extracting a portion sandwiched between long pause portions in the voice message. Means, utterance rate calculating means for calculating a syllable utterance rate in the recorded voice message, and extracting a part having a low utterance rate in the voice message, high-volume part calculating means, pause part calculating means, and uttering rate calculating means. Extraction part determining means for determining a part to be output as a part of the message from the part. . The high-volume portion calculating means assigns a weight to the extracted portion as a ratio to the average volume of the message, and the pause portion calculating portion weights the extracted portion as a ratio to the average value of the pause portion length in the message. The speech rate calculation means weights the ratio of the speech rate in the message to the average value to the extracted part, and the extraction part determination means includes the high volume part calculation means, the pause part calculation means, and the speech rate. A configuration may be adopted in which a predetermined number of parts are selected in the order in which the sum of the weights of the parts extracted by the calculation means becomes a large value, and these are set as parts output as part of the message.
Further, the vowel generating means for detecting a vowel generation timing in the message to determine a syllable delimiter is provided, and the utterance speed calculating means is configured to calculate the utterance speed from the number of syllables per unit time. Good.

[0008]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a recording / reproducing apparatus according to the present invention. As shown in FIG. 1, a recording / reproducing apparatus is a telephone receiving circuit or a PC.
Audio input means 1 such as an M receiving device, a tape recorder, a disk device, an audio recording means 2 such as a memory control device, a magnetic tape, a disk medium, an audio recording medium 3 such as a semiconductor memory, a tape recorder, a disk device, a memory control device And audio output means 5 such as a speaker.
including. If the sound is recorded as a digital signal,
The audio reproducing means 4 includes a DA converter.

The recording / reproducing apparatus includes a volume level detecting means 6 for detecting a volume level of a message recorded on the voice recording medium 3, a high volume portion calculating means 7 for calculating a high volume portion in the message, and a pause in the message. Pause part calculating means 8 for detecting a part, vowel detecting means 9 for detecting a vowel in a message, utterance speed calculating means 10 for calculating an utterance speed in a message, and extracted part determining means 1 for determining an important part in a message
1 is further provided. These volume level detecting means 6, high volume part calculating means 7, pause part calculating means 8, vowel detecting means 9, utterance speed calculating means 10, and extracted part determining means 11 are realized by software, for example. In this case, when audio is recorded as an analog signal on the audio recording medium 3, the audio signal is input to the software via the A / D converter.

Next, the operation will be described with reference to the flowchart of FIG. Generally speaking, when speaking, a speaker tends to speak an important part slowly and loudly between the front and back with a louder voice than other parts so as to be easy to hear. The recording / reproducing apparatus according to the present invention uses such a tendency to extract an important part in a message based on the volume, utterance speed and pause time of the recorded message.

The voice message input to the voice input means 1 is recorded on the voice recording medium 3 by the voice recording means 2. When the message is recorded, the message is sent to volume level detecting means 6 and vowel detecting means 9. The audio level detecting means 6 performs sampling and quantization, and calculates a relative audio intensity at each time from each quantized value (step S1). As a result, time-series sound intensity data is created. The sound intensity data is sent to the high sound volume part calculating means 7. The high-volume part calculating means 7 calculates a product of the sound intensities in a certain time unit (step S2). Then, a part having a large value is set as a part having a high volume in the whole, and is extracted as an important part candidate (step S3). The weight of the extracted part is given by the ratio of the volume of the part to the average volume of the message.

The volume intensity data calculated by the volume level detecting means 6 is sent to the pause portion calculating means 8. The pause part calculating means 8 calculates the product of the voice intensities in a unit of time of a certain length, and calculates a part having a small value as a pause part (step S4). In the calculated part,
The ratio of the length of the pause portion to the average value of the pause portion length in the message is given as a weight. Then, the length of each pause portion is compared, and a portion sandwiched between the long pause portions is extracted as an important portion candidate (step S5). Each candidate is given a weight as a percentage of the average length of the pause portion before and after the candidate with respect to the average value.

Japanese syllables are composed of vowels or consonants + vowels. Unless the same vowel continues twice, syllables are delimited at points of change between vowels and consonants and points of change to different vowels. Using this, the vowel detection means 9 divides the message into time units of a certain length shorter than one syllable (step S6). Then, the first to third formant frequencies are obtained for each of the divided regions using the zero-crossing number analysis. Further, by comparing each formant frequency in each region with the standard deviation of the vowel formant frequency, it is determined whether or not the region is a vowel (Step S).
7). Here, a range in which the same vowel continues is regarded as one vowel, which part is detected as a vowel in the whole, and a syllable is delimited (step S8). The data indicating the syllable delimiter in the message is output from the speech rate calculating means 10.
Sent to

The utterance speed calculating means 10 calculates the utterance speed based on the number of syllables per unit time (step S).
9) A part spoken at a low speed is calculated as an important part candidate (step S10). The weight of the extracted portion is given by the ratio of the speed of that portion to the average value of the utterance speed of the message.

The extracted part determining means 11 calculates the sum of the respective weights for the important part candidates determined by the high volume part calculating means 7, the pause part calculating means 8, and the speech speed calculating means 10. . For example, the high volume portion calculation means 7 sets A (2.0), B (2.5), and C (2.8) as candidates, and the pause portion calculation means 8 sets A (3.5), C (3.
0) and D (3.2) as candidates, and
Is a candidate for C (1.6) and D (1.5),
The sum of the weights of each part is A, B, C, D = (5.5),
(2.5), (7.4), and D (4.7). In addition,
Numerical values in parentheses show examples of weights.

Then, the extracted part determining means 11 finally determines some parts as important parts in the order of larger sum of weights (step S11). For example, a part having the largest sum of weights for one message is determined as an important part. Alternatively, two or three parts of one message are determined as important parts in descending order of the sum of weights. The part determined as described above is stored, for example, in a predetermined area in the audio recording medium 3.

Therefore, when the user plays the message,
First, by performing an operation of reproducing the important part determined by the extracted part determining means 11 first, it can be determined from the reproduced part whether the message including the part is a truly important message. If the user determines that the message is important, the entire message can be heard with priority. Although a configuration for performing such an operation is not shown in FIG. 1, a well-known configuration is disclosed in Japanese Patent Application Laid-Open Nos. 63-179652 and 3-186046. It is.

As described above, according to this embodiment,
Candidates include parts of the recorded message that are spoken larger than others, parts that are spoken slowly, and parts that are spoken with a gap between them. Is presented to the user as an important part, so that the user can be presented with a more reliable important part judgment material. Therefore, the user can hear the truly important message in the recorded message faster by reproducing the important part in advance. In the prior art, since the reliability of the important part judgment material is low, even when the entire message is heard based on such material, it is often not important. In other words, it took time to hear a truly important message. In this embodiment, important part candidates are extracted in consideration of three types of determination factors, that is, the volume, pause time, and utterance speed of a recorded message. Therefore, the possibility that the extracted part is an important part is not sufficient. high.

[0019]

As described above, according to the present invention, a recording / reproducing apparatus can be used in which a part of a message that is spoken larger than other parts, a part that is spoken more slowly, and a front and rear part are arranged. And the parts that are spoken are extracted, and some of them are configured as data that can be played back first, so that it is possible to know the importance of each message with high reliability first. This has the effect of enabling efficient message playback.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a recording / playback apparatus according to the present invention.

FIG. 2 is a flowchart showing the operation of the recording / playback apparatus according to the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 voice input means 2 voice recording means 3 voice recording medium 4 voice reproduction means 5 voice output means 6 volume level detection means 7 high volume portion calculation means 8 pause portion calculation means 9 vowel detection means 10 speech speed calculation means 11 extraction portion determination means

Claims (3)

[Claims] A recording / reproducing apparatus for recording an input voice message and outputting each voice message and a part of the voice message in response to a request from a user, wherein a high-volume portion of the voice message recorded is provided. A high volume portion calculating means for extracting a pause portion in the recorded voice message,
Pause portion calculation means for extracting a portion sandwiched between long pause portions in a voice message, utterance speed calculation means for calculating a syllable utterance speed in a recorded voice message, and extracting a portion having a low utterance speed in the voice message; A recording / reproducing apparatus, comprising: an extraction part determining means for determining a part to be output as a part of a message from the parts extracted by the high volume part calculating means, the pause part calculating means and the utterance speed calculating means. . 2. The high-volume-part calculating means assigns, as a weight, the ratio of the volume of the extracted part to the average value of the volume of the message to the extracted part, and the pause-part calculating means sandwiches the extracted long pause part. To the extracted portion, the weight of the ratio of the pause portion to the average value of the pause portion length in the message is given as a weight. The speech rate calculation means adds, to the extracted portion, the speech rate of the portion relative to the average value of the speech rate in the message. The proportion is given as a weight, and the extraction part determination means selects a predetermined number of parts in the order in which the sum of the weights of the parts extracted by the high volume part calculation means, the pause part calculation means and the speech speed calculation means becomes a large value. And
3. The recording / reproducing apparatus according to claim 2, wherein each selected part is a part output as a part of a message. 3. A vowel generating means for detecting a vowel generation timing in a message to determine a syllable delimiter, wherein the utterance speed calculating means calculates the utterance speed from the number of syllables per unit time. The recording / reproducing apparatus according to claim 2.